Compare commits

..

1 Commits

Author SHA1 Message Date
Sven Elfgren
5f745859c0 Started on md5 auth.
Left to figure out:
Whats the right format to store user:pw in userlist?
hashmap errors?
actually do hash:x comparison
2022-03-04 12:11:17 -07:00
42 changed files with 1537 additions and 4989 deletions

View File

@@ -9,20 +9,16 @@ jobs:
# Specify the execution environment. You can specify an image from Dockerhub or use one of our Convenience Images from CircleCI's Developer Hub.
# See: https://circleci.com/docs/2.0/configuration-reference/#docker-machine-macos-windows-executor
docker:
- image: levkk/pgcat-ci:latest
- image: cimg/rust:1.58.1
environment:
RUST_LOG: info
RUSTFLAGS: "-C instrument-coverage"
LLVM_PROFILE_FILE: "pgcat-%m.profraw"
- image: postgres:14
# auth:
# username: mydockerhub-user
# password: $DOCKERHUB_PASSWORD
- image: cimg/postgres:14.0
auth:
username: mydockerhub-user
password: $DOCKERHUB_PASSWORD
environment:
POSTGRES_USER: postgres
POSTGRES_DB: postgres
POSTGRES_PASSWORD: postgres
POSTGRES_HOST_AUTH_METHOD: scram-sha-256
# Add steps to the job
# See: https://circleci.com/docs/2.0/configuration-reference/#steps
steps:
@@ -34,19 +30,16 @@ jobs:
command: "cargo fmt --check"
- run:
name: "Install dependencies"
command: "sudo apt-get update && sudo apt-get install -y psmisc postgresql-contrib-12 postgresql-client-12 ruby ruby-dev libpq-dev python3 python3-pip lcov llvm-11 && sudo apt-get upgrade curl"
command: "sudo apt-get update && sudo apt-get install -y psmisc postgresql-contrib-12 postgresql-client-12 ruby ruby-dev libpq-dev"
- run:
name: "Install rust tools"
command: "cargo install cargo-binutils rustfilt && rustup component add llvm-tools-preview"
- run:
name: "Build"
name: "Build"
command: "cargo build"
- run:
name: "Tests"
command: "cargo test && bash .circleci/run_tests.sh && .circleci/generate_coverage.sh"
- store_artifacts:
path: /tmp/cov
destination: coverage-data
name: "Test"
command: "cargo test"
- run:
name: "Test end-to-end"
command: "bash .circleci/run_tests.sh"
- save_cache:
key: cargo-lock-2-{{ checksum "Cargo.lock" }}
paths:

View File

@@ -1,7 +0,0 @@
#!/bin/bash
rust-profdata merge -sparse pgcat-*.profraw -o pgcat.profdata
rust-cov export -ignore-filename-regex="rustc|registry" -Xdemangler=rustfilt -instr-profile=pgcat.profdata --object ./target/debug/pgcat --format lcov > ./lcov.info
genhtml lcov.info --output-directory /tmp/cov --prefix $(pwd)

View File

@@ -5,55 +5,78 @@
#
# General pooler settings
[general]
# What IP to run on, 0.0.0.0 means accessible from everywhere.
host = "0.0.0.0"
# Port to run on, same as PgBouncer used in this example.
port = 6432
# Whether to enable prometheus exporter or not.
enable_prometheus_exporter = true
# How many connections to allocate per server.
pool_size = 15
# Port at which prometheus exporter listens on.
prometheus_exporter_port = 9930
# How long to wait before aborting a server connection (ms).
connect_timeout = 100
# How much time to give the health check query to return with a result (ms).
healthcheck_timeout = 100
# How long to keep connection available for immediate re-use, without running a healthcheck query on it
healthcheck_delay = 30000
# How much time to give clients during shutdown before forcibly killing client connections (ms).
shutdown_timeout = 5000
# For how long to ban a server if it fails a health check (seconds).
ban_time = 60 # Seconds
# Reload config automatically if it changes.
autoreload = true
# TLS
tls_certificate = ".circleci/server.cert"
tls_private_key = ".circleci/server.key"
# Credentials to access the virtual administrative database (pgbouncer or pgcat)
# Connecting to that database allows running commands like `SHOW POOLS`, `SHOW DATABASES`, etc..
admin_username = "admin_user"
admin_password = "admin_pass"
# pool
# configs are structured as pool.<pool_name>
# the pool_name is what clients use as database name when connecting
# For the example below a client can connect using "postgres://sharding_user:sharding_user@pgcat_host:pgcat_port/sharded_db"
[pools.sharded_db]
# Pool mode (see PgBouncer docs for more).
# session: one server connection per connected client
# transaction: one server connection per client transaction
pool_mode = "transaction"
# How long to wait before aborting a server connection (ms).
connect_timeout = 100
# How much time to give `SELECT 1` health check query to return with a result (ms).
healthcheck_timeout = 100
# For how long to ban a server if it fails a health check (seconds).
ban_time = 60 # Seconds
# Stats will be sent here
statsd_address = "127.0.0.1:8125"
#
# User to use for authentication against the server.
[user]
name = "sharding_user"
password = "sharding_user"
#
# Shards in the cluster
[shards]
# Shard 0
[shards.0]
# [ host, port, role ]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5433, "replica" ],
# [ "127.0.1.1", 5432, "replica" ],
]
# Database name (e.g. "postgres")
database = "shard0"
[shards.1]
# [ host, port, role ]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5433, "replica" ],
# [ "127.0.1.1", 5432, "replica" ],
]
database = "shard1"
[shards.2]
# [ host, port, role ]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5433, "replica" ],
# [ "127.0.1.1", 5432, "replica" ],
]
database = "shard2"
# Settings for our query routing layer.
[query_router]
# If the client doesn't specify, route traffic to
# this role by default.
#
@@ -62,11 +85,12 @@ pool_mode = "transaction"
# primary: all queries go to the primary unless otherwise specified.
default_role = "any"
# Query parser. If enabled, we'll attempt to parse
# every incoming query to determine if it's a read or a write.
# If it's a read query, we'll direct it to a replica. Otherwise, if it's a write,
# we'll direct it to the primary.
query_parser_enabled = true
query_parser_enabled = false
# If the query parser is enabled and this setting is enabled, the primary will be part of the pool of databases used for
# load balancing of read queries. Otherwise, the primary will only be used for write
@@ -82,65 +106,3 @@ primary_reads_enabled = true
# sha1: A hashing function based on SHA1
#
sharding_function = "pg_bigint_hash"
# Credentials for users that may connect to this cluster
[pools.sharded_db.users.0]
username = "sharding_user"
password = "sharding_user"
# Maximum number of server connections that can be established for this user
# The maximum number of connection from a single Pgcat process to any database in the cluster
# is the sum of pool_size across all users.
pool_size = 9
statement_timeout = 0
[pools.sharded_db.users.1]
username = "other_user"
password = "other_user"
pool_size = 21
statement_timeout = 30000
# Shard 0
[pools.sharded_db.shards.0]
# [ host, port, role ]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ]
]
# Database name (e.g. "postgres")
database = "shard0"
search_path = "\"$user\",public"
[pools.sharded_db.shards.1]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ],
]
database = "shard1"
[pools.sharded_db.shards.2]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ],
]
database = "shard2"
[pools.simple_db]
pool_mode = "session"
default_role = "primary"
query_parser_enabled = true
primary_reads_enabled = true
sharding_function = "pg_bigint_hash"
[pools.simple_db.users.0]
username = "simple_user"
password = "simple_user"
pool_size = 5
statement_timeout = 30000
[pools.simple_db.shards.0]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ]
]
database = "some_db"

View File

@@ -3,9 +3,6 @@
set -e
set -o xtrace
# non-zero exit code if we provide bad configs
(! ./target/debug/pgcat "fake_configs" 2>/dev/null)
# Start PgCat with a particular log level
# for inspection.
function start_pgcat() {
@@ -15,15 +12,11 @@ function start_pgcat() {
}
# Setup the database with shards and user
PGPASSWORD=postgres psql -e -h 127.0.0.1 -p 5432 -U postgres -f tests/sharding/query_routing_setup.sql
PGPASSWORD=sharding_user pgbench -h 127.0.0.1 -U sharding_user shard0 -i
PGPASSWORD=sharding_user pgbench -h 127.0.0.1 -U sharding_user shard1 -i
PGPASSWORD=sharding_user pgbench -h 127.0.0.1 -U sharding_user shard2 -i
psql -e -h 127.0.0.1 -p 5432 -U postgres -f tests/sharding/query_routing_setup.sql
# Install Toxiproxy to simulate a downed/slow database
wget -O toxiproxy-2.4.0.deb https://github.com/Shopify/toxiproxy/releases/download/v2.4.0/toxiproxy_2.4.0_linux_$(dpkg --print-architecture).deb
sudo dpkg -i toxiproxy-2.4.0.deb
wget -O toxiproxy-2.1.4.deb https://github.com/Shopify/toxiproxy/releases/download/v2.1.4/toxiproxy_2.1.4_amd64.deb
sudo dpkg -i toxiproxy-2.1.4.deb
# Start Toxiproxy
toxiproxy-server &
@@ -34,93 +27,56 @@ toxiproxy-cli create -l 127.0.0.1:5433 -u 127.0.0.1:5432 postgres_replica
start_pgcat "info"
# Check that prometheus is running
curl --fail localhost:9930/metrics
export PGPASSWORD=sharding_user
export PGDATABASE=sharded_db
# pgbench test
pgbench -U sharding_user -i -h 127.0.0.1 -p 6432
pgbench -U sharding_user -h 127.0.0.1 -p 6432 -t 500 -c 2 --protocol simple -f tests/pgbench/simple.sql
pgbench -U sharding_user -h 127.0.0.1 -p 6432 -t 500 -c 2 --protocol extended
pgbench -i -h 127.0.0.1 -p 6432 && \
pgbench -h 127.0.0.1 -p 6432 -t 500 -c 2 --protocol simple && \
pgbench -h 127.0.0.1 -p 6432 -t 500 -c 2 --protocol extended
# COPY TO STDOUT test
psql -U sharding_user -h 127.0.0.1 -p 6432 -c 'COPY (SELECT * FROM pgbench_accounts LIMIT 15) TO STDOUT;' > /dev/null
psql -h 127.0.0.1 -p 6432 -c 'COPY (SELECT * FROM pgbench_accounts LIMIT 15) TO STDOUT;' > /dev/null
# Query cancellation test
(psql -U sharding_user -h 127.0.0.1 -p 6432 -c 'SELECT pg_sleep(50)' || true) &
sleep 1
killall psql -s SIGINT
# Reload pool (closing unused server connections)
PGPASSWORD=admin_pass psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'RELOAD'
(psql -U sharding_user -h 127.0.0.1 -p 6432 -c 'SELECT pg_sleep(50)' || true) &
sleep 1
(psql -h 127.0.0.1 -p 6432 -c 'SELECT pg_sleep(5)' || true) &
killall psql -s SIGINT
# Sharding insert
psql -U sharding_user -e -h 127.0.0.1 -p 6432 -f tests/sharding/query_routing_test_insert.sql
psql -e -h 127.0.0.1 -p 6432 -f tests/sharding/query_routing_test_insert.sql
# Sharding select
psql -U sharding_user -e -h 127.0.0.1 -p 6432 -f tests/sharding/query_routing_test_select.sql > /dev/null
psql -e -h 127.0.0.1 -p 6432 -f tests/sharding/query_routing_test_select.sql > /dev/null
# Replica/primary selection & more sharding tests
psql -U sharding_user -e -h 127.0.0.1 -p 6432 -f tests/sharding/query_routing_test_primary_replica.sql > /dev/null
# Statement timeout tests
sed -i 's/statement_timeout = 0/statement_timeout = 100/' .circleci/pgcat.toml
kill -SIGHUP $(pgrep pgcat) # Reload config
sleep 0.2
# This should timeout
(! psql -U sharding_user -e -h 127.0.0.1 -p 6432 -c 'select pg_sleep(0.5)')
# Disable statement timeout
sed -i 's/statement_timeout = 100/statement_timeout = 0/' .circleci/pgcat.toml
kill -SIGHUP $(pgrep pgcat) # Reload config again
psql -e -h 127.0.0.1 -p 6432 -f tests/sharding/query_routing_test_primary_replica.sql > /dev/null
#
# ActiveRecord tests
#
cd tests/ruby
sudo gem install bundler
bundle install
ruby tests.rb
cd tests/ruby && \
sudo gem install bundler && \
bundle install && \
ruby tests.rb && \
cd ../..
#
# Python tests
# These tests will start and stop the pgcat server so it will need to be restarted after the tests
#
pip3 install -r tests/python/requirements.txt
python3 tests/python/tests.py
start_pgcat "info"
# Admin tests
export PGPASSWORD=admin_pass
psql -U admin_user -e -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW STATS' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'RELOAD' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW CONFIG' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW DATABASES' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW LISTS' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW POOLS' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW VERSION' > /dev/null
psql -U admin_user -h 127.0.0.1 -p 6432 -d pgbouncer -c "SET client_encoding TO 'utf8'" > /dev/null # will ignore
(! psql -U admin_user -e -h 127.0.0.1 -p 6432 -d random_db -c 'SHOW STATS' > /dev/null)
export PGPASSWORD=sharding_user
psql -e -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW STATS' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'RELOAD' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW CONFIG' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW DATABASES' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW LISTS' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW POOLS' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW VERSION' > /dev/null
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c "SET client_encoding TO 'utf8'" > /dev/null # will ignore
(! psql -e -h 127.0.0.1 -p 6432 -d random_db -c 'SHOW STATS' > /dev/null)
# Start PgCat in debug to demonstrate failover better
start_pgcat "trace"
start_pgcat "debug"
# Add latency to the replica at port 5433 slightly above the healthcheck timeout
toxiproxy-cli toxic add -t latency -a latency=300 postgres_replica
sleep 1
# Note the failover in the logs
timeout 5 psql -U sharding_user -e -h 127.0.0.1 -p 6432 <<-EOF
timeout 5 psql -e -h 127.0.0.1 -p 6432 <<-EOF
SELECT 1;
SELECT 1;
SELECT 1;
@@ -132,18 +88,13 @@ toxiproxy-cli toxic remove --toxicName latency_downstream postgres_replica
start_pgcat "info"
# Test session mode (and config reload)
sed -i '0,/simple_db/s/pool_mode = "transaction"/pool_mode = "session"/' .circleci/pgcat.toml
sed -i 's/pool_mode = "transaction"/pool_mode = "session"/' pgcat.toml
# Reload config test
kill -SIGHUP $(pgrep pgcat)
# Revert settings after reload. Makes test runs idempotent
sed -i '0,/simple_db/s/pool_mode = "session"/pool_mode = "transaction"/' .circleci/pgcat.toml
sleep 1
# Prepared statements that will only work in session mode
pgbench -U sharding_user -h 127.0.0.1 -p 6432 -t 500 -c 2 --protocol prepared
pgbench -h 127.0.0.1 -p 6432 -t 500 -c 2 --protocol prepared
# Attempt clean shut down
killall pgcat -s SIGINT

View File

@@ -1,21 +0,0 @@
-----BEGIN CERTIFICATE-----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-----END CERTIFICATE-----

View File

@@ -1,28 +0,0 @@
-----BEGIN RSA PRIVATE KEY-----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-----END RSA PRIVATE KEY-----

5
.gitignore vendored
View File

@@ -1,4 +1,5 @@
.idea
/target
*.deb
.vscode
.idea/*
tests/ruby/.bundle/*
tests/ruby/vendor/*

View File

@@ -10,4 +10,10 @@ Happy hacking!
## TODOs
See [Issues]([url](https://github.com/levkk/pgcat/issues)).
A non-exhaustive list of things that would be useful to implement:
#### Client authentication
MD5 is probably sufficient, but maybe others too.
#### Admin
Admin database for stats collection and pooler administration. PgBouncer gives us a nice example on how to do that, specifically how to implement `RowDescription` and `DataRow` messages, [example here](https://github.com/pgbouncer/pgbouncer/blob/4f9ced8e63d317a6ff45c8b0efa876b32161f6db/src/admin.c#L813).

533
Cargo.lock generated
View File

@@ -45,12 +45,6 @@ version = "1.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cdb031dd78e28731d87d56cc8ffef4a8f36ca26c38fe2de700543e627f8a464a"
[[package]]
name = "base64"
version = "0.13.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "904dfeac50f3cdaba28fc6f57fdcddb75f49ed61346676a78c4ffe55877802fd"
[[package]]
name = "bb8"
version = "0.7.1"
@@ -79,24 +73,12 @@ dependencies = [
"generic-array",
]
[[package]]
name = "bumpalo"
version = "3.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "37ccbd214614c6783386c1af30caf03192f17891059cecc394b4fb119e363de3"
[[package]]
name = "bytes"
version = "1.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c4872d67bab6358e59559027aa3b9157c53d9358c51423c17554809a8858e0f8"
[[package]]
name = "cc"
version = "1.0.73"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2fff2a6927b3bb87f9595d67196a70493f627687a71d87a0d692242c33f58c11"
[[package]]
name = "cfg-if"
version = "1.0.0"
@@ -127,23 +109,22 @@ dependencies = [
[[package]]
name = "crypto-common"
version = "0.1.3"
version = "0.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "57952ca27b5e3606ff4dd79b0020231aaf9d6aa76dc05fd30137538c50bd3ce8"
checksum = "683d6b536309245c849479fba3da410962a43ed8e51c26b729208ec0ac2798d0"
dependencies = [
"generic-array",
"typenum",
]
[[package]]
name = "digest"
version = "0.10.3"
version = "0.10.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f2fb860ca6fafa5552fb6d0e816a69c8e49f0908bf524e30a90d97c85892d506"
checksum = "b697d66081d42af4fba142d56918a3cb21dc8eb63372c6b85d14f44fb9c5979b"
dependencies = [
"block-buffer",
"crypto-common",
"subtle",
"generic-array",
]
[[package]]
@@ -159,18 +140,6 @@ dependencies = [
"termcolor",
]
[[package]]
name = "exitcode"
version = "1.1.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "de853764b47027c2e862a995c34978ffa63c1501f2e15f987ba11bd4f9bba193"
[[package]]
name = "fnv"
version = "1.0.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
[[package]]
name = "futures-channel"
version = "0.3.19"
@@ -186,12 +155,6 @@ version = "0.3.19"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d0c8ff0461b82559810cdccfde3215c3f373807f5e5232b71479bff7bb2583d7"
[[package]]
name = "futures-sink"
version = "0.3.21"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "21163e139fa306126e6eedaf49ecdb4588f939600f0b1e770f4205ee4b7fa868"
[[package]]
name = "futures-task"
version = "0.3.19"
@@ -233,31 +196,6 @@ dependencies = [
"wasi",
]
[[package]]
name = "h2"
version = "0.3.13"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "37a82c6d637fc9515a4694bbf1cb2457b79d81ce52b3108bdeea58b07dd34a57"
dependencies = [
"bytes",
"fnv",
"futures-core",
"futures-sink",
"futures-util",
"http",
"indexmap",
"slab",
"tokio",
"tokio-util",
"tracing",
]
[[package]]
name = "hashbrown"
version = "0.12.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8a9ee70c43aaf417c914396645a0fa852624801b24ebb7ae78fe8272889ac888"
[[package]]
name = "hermit-abi"
version = "0.1.19"
@@ -267,89 +205,12 @@ dependencies = [
"libc",
]
[[package]]
name = "hmac"
version = "0.12.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6c49c37c09c17a53d937dfbb742eb3a961d65a994e6bcdcf37e7399d0cc8ab5e"
dependencies = [
"digest",
]
[[package]]
name = "http"
version = "0.2.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "75f43d41e26995c17e71ee126451dd3941010b0514a81a9d11f3b341debc2399"
dependencies = [
"bytes",
"fnv",
"itoa",
]
[[package]]
name = "http-body"
version = "0.4.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d5f38f16d184e36f2408a55281cd658ecbd3ca05cce6d6510a176eca393e26d1"
dependencies = [
"bytes",
"http",
"pin-project-lite",
]
[[package]]
name = "httparse"
version = "1.7.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "496ce29bb5a52785b44e0f7ca2847ae0bb839c9bd28f69acac9b99d461c0c04c"
[[package]]
name = "httpdate"
version = "1.0.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c4a1e36c821dbe04574f602848a19f742f4fb3c98d40449f11bcad18d6b17421"
[[package]]
name = "humantime"
version = "2.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9a3a5bfb195931eeb336b2a7b4d761daec841b97f947d34394601737a7bba5e4"
[[package]]
name = "hyper"
version = "0.14.20"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "02c929dc5c39e335a03c405292728118860721b10190d98c2a0f0efd5baafbac"
dependencies = [
"bytes",
"futures-channel",
"futures-core",
"futures-util",
"h2",
"http",
"http-body",
"httparse",
"httpdate",
"itoa",
"pin-project-lite",
"socket2",
"tokio",
"tower-service",
"tracing",
"want",
]
[[package]]
name = "indexmap"
version = "1.9.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "10a35a97730320ffe8e2d410b5d3b69279b98d2c14bdb8b70ea89ecf7888d41e"
dependencies = [
"autocfg",
"hashbrown",
]
[[package]]
name = "instant"
version = "0.1.12"
@@ -361,30 +222,15 @@ dependencies = [
[[package]]
name = "itoa"
version = "1.0.3"
version = "1.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6c8af84674fe1f223a982c933a0ee1086ac4d4052aa0fb8060c12c6ad838e754"
[[package]]
name = "js-sys"
version = "0.3.58"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c3fac17f7123a73ca62df411b1bf727ccc805daa070338fda671c86dac1bdc27"
dependencies = [
"wasm-bindgen",
]
[[package]]
name = "lazy_static"
version = "1.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e2abad23fbc42b3700f2f279844dc832adb2b2eb069b2df918f455c4e18cc646"
checksum = "1aab8fc367588b89dcee83ab0fd66b72b50b72fa1904d7095045ace2b0c81c35"
[[package]]
name = "libc"
version = "0.2.126"
version = "0.2.117"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "349d5a591cd28b49e1d1037471617a32ddcda5731b99419008085f72d5a53836"
checksum = "e74d72e0f9b65b5b4ca49a346af3976df0f9c61d550727f349ecd559f251a26c"
[[package]]
name = "lock_api"
@@ -512,82 +358,31 @@ dependencies = [
[[package]]
name = "pgcat"
version = "0.6.0-alpha1"
version = "0.1.0"
dependencies = [
"arc-swap",
"async-trait",
"base64",
"bb8",
"bytes",
"chrono",
"env_logger",
"exitcode",
"hmac",
"hyper",
"log",
"md-5",
"num_cpus",
"once_cell",
"parking_lot",
"phf",
"rand",
"regex",
"rustls-pemfile",
"serde",
"serde_derive",
"serde_json",
"sha-1",
"sha2",
"sqlparser",
"stringprep",
"statsd",
"tokio",
"tokio-rustls",
"toml",
]
[[package]]
name = "phf"
version = "0.10.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fabbf1ead8a5bcbc20f5f8b939ee3f5b0f6f281b6ad3468b84656b658b455259"
dependencies = [
"phf_macros",
"phf_shared",
"proc-macro-hack",
]
[[package]]
name = "phf_generator"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5d5285893bb5eb82e6aaf5d59ee909a06a16737a8970984dd7746ba9283498d6"
dependencies = [
"phf_shared",
"rand",
]
[[package]]
name = "phf_macros"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "58fdf3184dd560f160dd73922bea2d5cd6e8f064bf4b13110abd81b03697b4e0"
dependencies = [
"phf_generator",
"phf_shared",
"proc-macro-hack",
"proc-macro2",
"quote",
"syn",
]
[[package]]
name = "phf_shared"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b6796ad771acdc0123d2a88dc428b5e38ef24456743ddb1744ed628f9815c096"
dependencies = [
"siphasher",
]
[[package]]
name = "pin-project-lite"
version = "0.2.8"
@@ -606,12 +401,6 @@ version = "0.2.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eb9f9e6e233e5c4a35559a617bf40a4ec447db2e84c20b55a6f83167b7e57872"
[[package]]
name = "proc-macro-hack"
version = "0.5.19"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "dbf0c48bc1d91375ae5c3cd81e3722dff1abcf81a30960240640d223f59fe0e5"
[[package]]
name = "proc-macro2"
version = "1.0.36"
@@ -681,9 +470,9 @@ dependencies = [
[[package]]
name = "regex"
version = "1.5.5"
version = "1.5.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1a11647b6b25ff05a515cb92c365cec08801e83423a235b51e231e1808747286"
checksum = "d07a8629359eb56f1e2fb1652bb04212c072a87ba68546a04065d525673ac461"
dependencies = [
"aho-corasick",
"memchr",
@@ -697,40 +486,10 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f497285884f3fcff424ffc933e56d7cbca511def0c9831a7f9b5f6153e3cc89b"
[[package]]
name = "ring"
version = "0.16.20"
name = "ryu"
version = "1.0.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3053cf52e236a3ed746dfc745aa9cacf1b791d846bdaf412f60a8d7d6e17c8fc"
dependencies = [
"cc",
"libc",
"once_cell",
"spin",
"untrusted",
"web-sys",
"winapi",
]
[[package]]
name = "rustls"
version = "0.20.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5aab8ee6c7097ed6057f43c187a62418d0c05a4bd5f18b3571db50ee0f9ce033"
dependencies = [
"log",
"ring",
"sct",
"webpki",
]
[[package]]
name = "rustls-pemfile"
version = "1.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e7522c9de787ff061458fe9a829dc790a3f5b22dc571694fc5883f448b94d9a9"
dependencies = [
"base64",
]
checksum = "73b4b750c782965c211b42f022f59af1fbceabdd026623714f104152f1ec149f"
[[package]]
name = "scopeguard"
@@ -738,16 +497,6 @@ version = "1.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d29ab0c6d3fc0ee92fe66e2d99f700eab17a8d57d1c1d3b748380fb20baa78cd"
[[package]]
name = "sct"
version = "0.7.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d53dcdb7c9f8158937a7981b48accfd39a43af418591a5d008c7b22b5e1b7ca4"
dependencies = [
"ring",
"untrusted",
]
[[package]]
name = "serde"
version = "1.0.136"
@@ -765,6 +514,17 @@ dependencies = [
"syn",
]
[[package]]
name = "serde_json"
version = "1.0.79"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8e8d9fa5c3b304765ce1fd9c4c8a3de2c8db365a5b91be52f186efc675681d95"
dependencies = [
"itoa",
"ryu",
"serde",
]
[[package]]
name = "sha-1"
version = "0.10.0"
@@ -776,17 +536,6 @@ dependencies = [
"digest",
]
[[package]]
name = "sha2"
version = "0.10.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "55deaec60f81eefe3cce0dc50bda92d6d8e88f2a27df7c5033b42afeb1ed2676"
dependencies = [
"cfg-if",
"cpufeatures",
"digest",
]
[[package]]
name = "signal-hook-registry"
version = "1.4.0"
@@ -796,12 +545,6 @@ dependencies = [
"libc",
]
[[package]]
name = "siphasher"
version = "0.3.10"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7bd3e3206899af3f8b12af284fafc038cc1dc2b41d1b89dd17297221c5d225de"
[[package]]
name = "slab"
version = "0.4.5"
@@ -814,22 +557,6 @@ version = "1.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f2dd574626839106c320a323308629dcb1acfc96e32a8cba364ddc61ac23ee83"
[[package]]
name = "socket2"
version = "0.4.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "66d72b759436ae32898a2af0a14218dbf55efde3feeb170eb623637db85ee1e0"
dependencies = [
"libc",
"winapi",
]
[[package]]
name = "spin"
version = "0.5.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6e63cff320ae2c57904679ba7cb63280a3dc4613885beafb148ee7bf9aa9042d"
[[package]]
name = "sqlparser"
version = "0.14.0"
@@ -840,21 +567,14 @@ dependencies = [
]
[[package]]
name = "stringprep"
version = "0.1.2"
name = "statsd"
version = "0.15.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8ee348cb74b87454fff4b551cbf727025810a004f88aeacae7f85b87f4e9a1c1"
checksum = "df1efceb4bf2c0b5ebec94354285a43bbbed1375605bdf2ebe4132299434a330"
dependencies = [
"unicode-bidi",
"unicode-normalization",
"rand",
]
[[package]]
name = "subtle"
version = "2.4.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6bdef32e8150c2a081110b42772ffe7d7c9032b606bc226c8260fd97e0976601"
[[package]]
name = "syn"
version = "1.0.86"
@@ -886,21 +606,6 @@ dependencies = [
"winapi",
]
[[package]]
name = "tinyvec"
version = "1.6.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "87cc5ceb3875bb20c2890005a4e226a4651264a5c75edb2421b52861a0a0cb50"
dependencies = [
"tinyvec_macros",
]
[[package]]
name = "tinyvec_macros"
version = "0.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "cda74da7e1a664f795bb1f8a87ec406fb89a02522cf6e50620d016add6dbbf5c"
[[package]]
name = "tokio"
version = "1.16.1"
@@ -931,31 +636,6 @@ dependencies = [
"syn",
]
[[package]]
name = "tokio-rustls"
version = "0.23.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c43ee83903113e03984cb9e5cebe6c04a5116269e900e3ddba8f068a62adda59"
dependencies = [
"rustls",
"tokio",
"webpki",
]
[[package]]
name = "tokio-util"
version = "0.7.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f988a1a1adc2fb21f9c12aa96441da33a1728193ae0b95d2be22dbd17fcb4e5c"
dependencies = [
"bytes",
"futures-core",
"futures-sink",
"pin-project-lite",
"tokio",
"tracing",
]
[[package]]
name = "toml"
version = "0.5.8"
@@ -965,179 +645,30 @@ dependencies = [
"serde",
]
[[package]]
name = "tower-service"
version = "0.3.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b6bc1c9ce2b5135ac7f93c72918fc37feb872bdc6a5533a8b85eb4b86bfdae52"
[[package]]
name = "tracing"
version = "0.1.34"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5d0ecdcb44a79f0fe9844f0c4f33a342cbcbb5117de8001e6ba0dc2351327d09"
dependencies = [
"cfg-if",
"pin-project-lite",
"tracing-attributes",
"tracing-core",
]
[[package]]
name = "tracing-attributes"
version = "0.1.22"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "11c75893af559bc8e10716548bdef5cb2b983f8e637db9d0e15126b61b484ee2"
dependencies = [
"proc-macro2",
"quote",
"syn",
]
[[package]]
name = "tracing-core"
version = "0.1.26"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f54c8ca710e81886d498c2fd3331b56c93aa248d49de2222ad2742247c60072f"
dependencies = [
"lazy_static",
]
[[package]]
name = "try-lock"
version = "0.2.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "59547bce71d9c38b83d9c0e92b6066c4253371f15005def0c30d9657f50c7642"
[[package]]
name = "typenum"
version = "1.15.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "dcf81ac59edc17cc8697ff311e8f5ef2d99fcbd9817b34cec66f90b6c3dfd987"
[[package]]
name = "unicode-bidi"
version = "0.3.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "099b7128301d285f79ddd55b9a83d5e6b9e97c92e0ea0daebee7263e932de992"
[[package]]
name = "unicode-normalization"
version = "0.1.19"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d54590932941a9e9266f0832deed84ebe1bf2e4c9e4a3554d393d18f5e854bf9"
dependencies = [
"tinyvec",
]
[[package]]
name = "unicode-xid"
version = "0.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8ccb82d61f80a663efe1f787a51b16b5a51e3314d6ac365b08639f52387b33f3"
[[package]]
name = "untrusted"
version = "0.7.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a156c684c91ea7d62626509bce3cb4e1d9ed5c4d978f7b4352658f96a4c26b4a"
[[package]]
name = "version_check"
version = "0.9.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f"
[[package]]
name = "want"
version = "0.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1ce8a968cb1cd110d136ff8b819a556d6fb6d919363c61534f6860c7eb172ba0"
dependencies = [
"log",
"try-lock",
]
[[package]]
name = "wasi"
version = "0.10.0+wasi-snapshot-preview1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1a143597ca7c7793eff794def352d41792a93c481eb1042423ff7ff72ba2c31f"
[[package]]
name = "wasm-bindgen"
version = "0.2.81"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7c53b543413a17a202f4be280a7e5c62a1c69345f5de525ee64f8cfdbc954994"
dependencies = [
"cfg-if",
"wasm-bindgen-macro",
]
[[package]]
name = "wasm-bindgen-backend"
version = "0.2.81"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5491a68ab4500fa6b4d726bd67408630c3dbe9c4fe7bda16d5c82a1fd8c7340a"
dependencies = [
"bumpalo",
"lazy_static",
"log",
"proc-macro2",
"quote",
"syn",
"wasm-bindgen-shared",
]
[[package]]
name = "wasm-bindgen-macro"
version = "0.2.81"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c441e177922bc58f1e12c022624b6216378e5febc2f0533e41ba443d505b80aa"
dependencies = [
"quote",
"wasm-bindgen-macro-support",
]
[[package]]
name = "wasm-bindgen-macro-support"
version = "0.2.81"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7d94ac45fcf608c1f45ef53e748d35660f168490c10b23704c7779ab8f5c3048"
dependencies = [
"proc-macro2",
"quote",
"syn",
"wasm-bindgen-backend",
"wasm-bindgen-shared",
]
[[package]]
name = "wasm-bindgen-shared"
version = "0.2.81"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6a89911bd99e5f3659ec4acf9c4d93b0a90fe4a2a11f15328472058edc5261be"
[[package]]
name = "web-sys"
version = "0.3.58"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2fed94beee57daf8dd7d51f2b15dc2bcde92d7a72304cdf662a4371008b71b90"
dependencies = [
"js-sys",
"wasm-bindgen",
]
[[package]]
name = "webpki"
version = "0.22.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f095d78192e208183081cc07bc5515ef55216397af48b873e5edcd72637fa1bd"
dependencies = [
"ring",
"untrusted",
]
[[package]]
name = "winapi"
version = "0.3.9"

View File

@@ -1,6 +1,6 @@
[package]
name = "pgcat"
version = "0.6.0-alpha1"
version = "0.1.0"
edition = "2021"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
@@ -17,20 +17,13 @@ sha-1 = "0.10"
toml = "0.5"
serde = "1"
serde_derive = "1"
serde_json = "1"
regex = "1"
num_cpus = "1"
once_cell = "1"
statsd = "0.15"
sqlparser = "0.14"
log = "0.4"
arc-swap = "1"
env_logger = "0.9"
parking_lot = "0.11"
hmac = "0.12"
sha2 = "0.10"
base64 = "0.13"
stringprep = "0.1"
tokio-rustls = "0.23"
rustls-pemfile = "1"
hyper = { version = "0.14", features = ["full"] }
phf = { version = "0.10", features = ["macros"] }
exitcode = "1.1.2"

View File

@@ -1,11 +1,11 @@
FROM rust:1 AS builder
FROM rust:1.58-slim-buster AS builder
COPY . /app
WORKDIR /app
RUN cargo build --release
FROM debian:bullseye-slim
FROM debian:buster-slim
COPY --from=builder /app/target/release/pgcat /usr/bin/pgcat
COPY --from=builder /app/pgcat.toml /etc/pgcat/pgcat.toml
WORKDIR /etc/pgcat
ENV RUST_LOG=info
CMD ["pgcat"]
ENTRYPOINT ["/usr/bin/pgcat"]

View File

@@ -1,8 +0,0 @@
FROM cimg/rust:1.62.0
RUN sudo apt-get update && \
sudo apt-get install -y psmisc postgresql-contrib-12 postgresql-client-12 ruby ruby-dev libpq-dev python3 python3-pip lcov llvm-11 && \
sudo apt-get upgrade curl
RUN cargo install cargo-binutils rustfilt && \
rustup component add llvm-tools-preview
RUN pip3 install psycopg2 && \
sudo gem install bundler

125
README.md
View File

@@ -4,68 +4,54 @@
![PgCat](./pgcat3.png)
PostgreSQL pooler (like PgBouncer) with sharding, load balancing and failover support.
Meow. PgBouncer rewritten in Rust, with sharding, load balancing and failover support.
**Beta**: looking for beta testers, see [#35](https://github.com/levkk/pgcat/issues/35).
**Alpha**: looking for alpha testers, see [#35](https://github.com/levkk/pgcat/issues/35).
## Features
| **Feature** | **Status** | **Comments** |
|--------------------------------|-----------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------|
| Transaction pooling | :white_check_mark: | Identical to PgBouncer. |
| Session pooling | :white_check_mark: | Identical to PgBouncer. |
| `COPY` support | :white_check_mark: | Both `COPY TO` and `COPY FROM` are supported. |
| Query cancellation | :white_check_mark: | Supported both in transaction and session pooling modes. |
| Load balancing of read queries | :white_check_mark: | Using random between replicas. Primary is included when `primary_reads_enabled` is enabled (default). |
| Sharding | :white_check_mark: | Transactions are sharded using `SET SHARD TO` and `SET SHARDING KEY TO` syntax extensions; see examples below. |
| Failover | :white_check_mark: | Replicas are tested with a health check. If a health check fails, remaining replicas are attempted; see below for algorithm description and examples. |
| Statistics | :white_check_mark: | Statistics available in the admin database (`pgcat` and `pgbouncer`) with `SHOW STATS`, `SHOW POOLS` and others. |
| Live configuration reloading | :white_check_mark: | Reload supported settings with a `SIGHUP` to the process, e.g. `kill -s SIGHUP $(pgrep pgcat)` or `RELOAD` query issued to the admin database. |
| Client authentication | :white_check_mark: :wrench: | MD5 password authentication is supported, SCRAM is on the roadmap; one user is used to connect to Postgres with both SCRAM and MD5 supported. |
| Admin database | :white_check_mark: | The admin database, similar to PgBouncer's, allows to query for statistics and reload the configuration. |
| **Feature** | **Status** | **Comments** |
|--------------------------------|-----------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------|
| Transaction pooling | :white_check_mark: | Identical to PgBouncer. |
| Session pooling | :white_check_mark: | Identical to PgBouncer. |
| `COPY` support | :white_check_mark: | Both `COPY TO` and `COPY FROM` are supported. |
| Query cancellation | :white_check_mark: | Supported both in transaction and session pooling modes. |
| Load balancing of read queries | :white_check_mark: | Using round-robin between replicas. Primary is included when `primary_reads_enabled` is enabled (default). |
| Sharding | :white_check_mark: | Transactions are sharded using `SET SHARD TO` and `SET SHARDING KEY TO` syntax extensions; see examples below. |
| Failover | :white_check_mark: | Replicas are tested with a health check. If a health check fails, remaining replicas are attempted; see below for algorithm description and examples. |
| Statistics reporting | :white_check_mark: | Statistics similar to PgBouncers are reported via StatsD. |
| Live configuration reloading | :construction_worker: | Reload config with a `SIGHUP` to the process, e.g. `kill -s SIGHUP $(pgrep pgcat)`. Not all settings can be reloaded without a restart. |
| Client authentication | :x: :wrench: | On the roadmap; currently all clients are allowed to connect and one user is used to connect to Postgres. |
## Deployment
See `Dockerfile` for example deployment using Docker. The pooler is configured to spawn 4 workers so 4 CPUs are recommended for optimal performance. That setting can be adjusted to spawn as many (or as little) workers as needed.
For quick local example, use the Docker Compose environment provided:
```bash
docker-compose up
# In a new terminal:
psql -h 127.0.0.1 -p 6432 -c 'SELECT 1'
```
See `Dockerfile` for example deployment using Docker. The pooler is configured to spawn 4 workers so 4 CPUs are recommended for optimal performance.
That setting can be adjusted to spawn as many (or as little) workers as needed.
### Config
| **Name** | **Description** | **Examples** |
|------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------|
| **`general`** | | |
| `host` | The pooler will run on this host, 0.0.0.0 means accessible from everywhere. | `0.0.0.0` |
| `port` | The pooler will run on this port. | `6432` |
| `enable_prometheus_exporter` | Enable prometheus exporter which will export metrics in prometheus exposition format. | `true` |
| `prometheus_exporter_port` | Port at which prometheus exporter listens on. | `9930` |
| `pool_size` | Maximum allowed server connections per pool. Pools are separated for each user/shard/server role. The connections are allocated as needed. | `15` |
| `pool_mode` | The pool mode to use, i.e. `session` or `transaction`. | `transaction` |
| `connect_timeout` | Maximum time to establish a connection to a server (milliseconds). If reached, the server is banned and the next target is attempted. | `5000` |
| `healthcheck_timeout` | Maximum time to pass a health check (`SELECT 1`, milliseconds). If reached, the server is banned and the next target is attempted. | `1000` |
| `shutdown_timeout` | Maximum time to give clients during shutdown before forcibly killing client connections (ms). | `60000` |
| `healthcheck_delay` | How long to keep connection available for immediate re-use, without running a healthcheck query on it | `30000` |
| `ban_time` | Ban time for a server (seconds). It won't be allowed to serve transactions until the ban expires; failover targets will be used instead. | `60` |
| `autoreload` | Enable auto-reload of config after fixed time-interval. | `false` |
| | | |
| **`user`** | | |
| `name` | The user name. | `sharding_user` |
| `password` | The user password in plaintext. | `hunter2` |
| | | |
| **`shards`** | Shards are numerically numbered starting from 0; the order in the config is preserved by the pooler to route queries accordingly. | `[shards.0]` |
| `servers` | List of servers to connect to and their roles. A server is: `[host, port, role]`, where `role` is either `primary` or `replica`. | `["127.0.0.1", 5432, "primary"]` |
| `database` | The name of the database to connect to. This is the same on all servers that are part of one shard. | |
| | | |
| **`query_router`** | | |
| `default_role` | Traffic is routed to this role by default (random), unless the client specifies otherwise. Default is `any`, for any role available. | `any`, `primary`, `replica` |
| `query_parser_enabled` | Enable the query parser which will inspect incoming queries and route them to a primary or replicas. | `false` |
| `primary_reads_enabled` | Enable this to allow read queries on the primary; otherwise read queries are routed to the replicas. | `true` |
| **Name** | **Description** | **Examples** |
|-------------------------|--------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------|
| **`general`** | | |
| `host` | The pooler will run on this host, 0.0.0.0 means accessible from everywhere. | `0.0.0.0` |
| `port` | The pooler will run on this port. | `6432` |
| `pool_size` | Maximum allowed server connections per pool. Pools are separated for each user/shard/server role. The connections are allocated as needed. | `15` |
| `pool_mode` | The pool mode to use, i.e. `session` or `transaction`. | `transaction` |
| `connect_timeout` | Maximum time to establish a connection to a server (milliseconds). If reached, the server is banned and the next target is attempted. | `5000` |
| `healthcheck_timeout` | Maximum time to pass a health check (`SELECT 1`, milliseconds). If reached, the server is banned and the next target is attempted. | `1000` |
| `ban_time` | Ban time for a server (seconds). It won't be allowed to serve transactions until the ban expires; failover targets will be used instead. | `60` |
| `statsd_address` | StatsD host and port. Statistics will be sent there every 15 seconds. | `127.0.0.1:8125` |
| | | |
| **`user`** | | |
| `name` | The user name. | `sharding_user` |
| `password` | The user password in plaintext. | `hunter2` |
| | | |
| **`shards`** | Shards are numerically numbered starting from 0; the order in the config is preserved by the pooler to route queries accordingly. | `[shards.0]` |
| `servers` | List of servers to connect to and their roles. A server is: `[host, port, role]`, where `role` is either `primary` or `replica`. | `["127.0.0.1", 5432, "primary"]` |
| `database` | The name of the database to connect to. This is the same on all servers that are part of one shard. | |
| **`query_router`** | | |
| `default_role` | Traffic is routed to this role by default (round-robin), unless the client specifies otherwise. Default is `any`, for any role available. | `any`, `primary`, `replica` |
| `query_parser_enabled` | Enable the query parser which will inspect incoming queries and route them to a primary or replicas. | `false` |
| `primary_reads_enabled` | Enable this to allow read queries on the primary; otherwise read queries are routed to the replicas. | `true` |
## Local development
@@ -96,7 +82,7 @@ See [sharding README](./tests/sharding/README.md) for sharding logic testing.
| Load balancing | :white_check_mark: | :white_check_mark: | We could test this by emitting statistics for each replica and compare them. |
| Failover | :white_check_mark: | :white_check_mark: | Misconfigure a replica in `pgcat.toml` and watch it forward queries to spares. CI testing is using Toxiproxy. |
| Sharding | :white_check_mark: | :white_check_mark: | See `tests/sharding` and `tests/ruby` for an Rails/ActiveRecord example. |
| Statistics | :white_check_mark: | :white_check_mark: | Query the admin database with `psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW STATS'`. |
| Statistics reporting | :x: | :white_check_mark: | Run `nc -l -u 8125` and watch the stats come in every 15 seconds. |
| Live config reloading | :white_check_mark: | :white_check_mark: | Run `kill -s SIGHUP $(pgrep pgcat)` and watch the config reload. |
## Usage
@@ -112,7 +98,7 @@ In transaction mode, a client talks to one server for the duration of a single t
This mode is enabled by default.
### Load balancing of read queries
All queries are load balanced against the configured servers using the random algorithm. The most straight forward configuration example would be to put this pooler in front of several replicas and let it load balance all queries.
All queries are load balanced against the configured servers using the round-robin algorithm. The most straight forward configuration example would be to put this pooler in front of several replicas and let it load balance all queries.
If the configuration includes a primary and replicas, the queries can be separated with the built-in query parser. The query parser will interpret the query and route all `SELECT` queries to a replica, while all other queries including explicit transactions will be routed to the primary.
@@ -151,18 +137,18 @@ Failover behavior can get pretty interesting (read complex) when multiple config
| **Query** | **`SET SERVER ROLE TO`** | **`query_parser_enabled`** | **`primary_reads_enabled`** | **Target state** | **Outcome** |
|---------------------------|--------------------------|----------------------------|-----------------------------|------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Read query, i.e. `SELECT` | unset (any) | false | false | up | Query is routed to the first instance in the random loop. |
| Read query | unset (any) | true | false | up | Query is routed to the first replica instance in the random loop. |
| Read query | unset (any) | true | true | up | Query is routed to the first instance in the random loop. |
| Read query | replica | false | false | up | Query is routed to the first replica instance in the random loop. |
| Read query, i.e. `SELECT` | unset (any) | false | false | up | Query is routed to the first instance in the round-robin loop. |
| Read query | unset (any) | true | false | up | Query is routed to the first replica instance in the round-robin loop. |
| Read query | unset (any) | true | true | up | Query is routed to the first instance in the round-robin loop. |
| Read query | replica | false | false | up | Query is routed to the first replica instance in the round-robin loop. |
| Read query | primary | false | false | up | Query is routed to the primary. |
| Read query | unset (any) | false | false | down | First instance is banned for reads. Next target in the random loop is attempted. |
| Read query | unset (any) | true | false | down | First replica instance is banned. Next replica instance is attempted in the random loop. |
| Read query | unset (any) | true | true | down | First instance (even if primary) is banned for reads. Next instance is attempted in the random loop. |
| Read query | replica | false | false | down | First replica instance is banned. Next replica instance is attempted in the random loop. |
| Read query | unset (any) | false | false | down | First instance is banned for reads. Next target in the round-robin loop is attempted. |
| Read query | unset (any) | true | false | down | First replica instance is banned. Next replica instance is attempted in the round-robin loop. |
| Read query | unset (any) | true | true | down | First instance (even if primary) is banned for reads. Next instance is attempted in the round-robin loop. |
| Read query | replica | false | false | down | First replica instance is banned. Next replica instance is attempted in the round-robin loop. |
| Read query | primary | false | false | down | The query is attempted against the primary and fails. The client receives an error. |
| | | | | | |
| Write query e.g. `INSERT` | unset (any) | false | false | up | The query is attempted against the first available instance in the random loop. If the instance is a replica, the query fails and the client receives an error. |
| Write query e.g. `INSERT` | unset (any) | false | false | up | The query is attempted against the first available instance in the round-robin loop. If the instance is a replica, the query fails and the client receives an error. |
| Write query | unset (any) | true | false | up | The query is routed to the primary. |
| Write query | unset (any) | true | true | up | The query is routed to the primary. |
| Write query | primary | false | false | up | The query is routed to the primary. |
@@ -239,15 +225,11 @@ SELECT * FROM users WHERE email = 'test@example.com'; -- shard setting lasts unt
### Statistics reporting
The stats are very similar to what Pgbouncer reports and the names are kept to be comparable. They are accessible by querying the admin database `pgcat`, and `pgbouncer` for compatibility.
```
psql -h 127.0.0.1 -p 6432 -d pgbouncer -c 'SHOW DATABASES'
```
Stats are reported using StatsD every 15 seconds. The address is configurable with `statsd_address`, the default is `127.0.0.1:8125`. The stats are very similar to what Pgbouncer reports and the names are kept to be comparable.
### Live configuration reloading
The config can be reloaded by sending a `kill -s SIGHUP` to the process or by querying `RELOAD` to the admin database. Not all settings are currently supported by live reload:
The config can be reloaded by sending a `kill -s SIGHUP` to the process. Not all settings are currently supported by live reload:
| **Config** | **Requires restart** |
|-------------------------|----------------------|
@@ -256,9 +238,8 @@ The config can be reloaded by sending a `kill -s SIGHUP` to the process or by qu
| `pool_mode` | no |
| `connect_timeout` | yes |
| `healthcheck_timeout` | no |
| `shutdown_timeout` | no |
| `healthcheck_delay` | no |
| `ban_time` | no |
| `statsd_address` | yes |
| `user` | yes |
| `shards` | yes |
| `default_role` | no |

View File

@@ -1,17 +0,0 @@
version: "3"
services:
postgres:
image: postgres:14
environment:
POSTGRES_PASSWORD: postgres
POSTGRES_HOST_AUTH_METHOD: md5
pgcat:
build: .
command:
- "pgcat"
- "/etc/pgcat/pgcat.toml"
volumes:
- "${PWD}/examples/docker/pgcat.toml:/etc/pgcat/pgcat.toml"
ports:
- "6432:6432"
- "9930:9930"

View File

@@ -1,147 +0,0 @@
#
# PgCat config example.
#
#
# General pooler settings
[general]
# What IP to run on, 0.0.0.0 means accessible from everywhere.
host = "0.0.0.0"
# Port to run on, same as PgBouncer used in this example.
port = 6432
# Whether to enable prometheus exporter or not.
enable_prometheus_exporter = true
# Port at which prometheus exporter listens on.
prometheus_exporter_port = 9930
# How long to wait before aborting a server connection (ms).
connect_timeout = 5000
# How much time to give `SELECT 1` health check query to return with a result (ms).
healthcheck_timeout = 1000
# How long to keep connection available for immediate re-use, without running a healthcheck query on it
healthcheck_delay = 30000
# How much time to give clients during shutdown before forcibly killing client connections (ms).
shutdown_timeout = 60000
# For how long to ban a server if it fails a health check (seconds).
ban_time = 60 # seconds
# Reload config automatically if it changes.
autoreload = false
# TLS
# tls_certificate = "server.cert"
# tls_private_key = "server.key"
# Credentials to access the virtual administrative database (pgbouncer or pgcat)
# Connecting to that database allows running commands like `SHOW POOLS`, `SHOW DATABASES`, etc..
admin_username = "postgres"
admin_password = "postgres"
# pool
# configs are structured as pool.<pool_name>
# the pool_name is what clients use as database name when connecting
# For the example below a client can connect using "postgres://sharding_user:sharding_user@pgcat_host:pgcat_port/sharded"
[pools.sharded]
# Pool mode (see PgBouncer docs for more).
# session: one server connection per connected client
# transaction: one server connection per client transaction
pool_mode = "transaction"
# If the client doesn't specify, route traffic to
# this role by default.
#
# any: round-robin between primary and replicas,
# replica: round-robin between replicas only without touching the primary,
# primary: all queries go to the primary unless otherwise specified.
default_role = "any"
# Query parser. If enabled, we'll attempt to parse
# every incoming query to determine if it's a read or a write.
# If it's a read query, we'll direct it to a replica. Otherwise, if it's a write,
# we'll direct it to the primary.
query_parser_enabled = true
# If the query parser is enabled and this setting is enabled, the primary will be part of the pool of databases used for
# load balancing of read queries. Otherwise, the primary will only be used for write
# queries. The primary can always be explicitely selected with our custom protocol.
primary_reads_enabled = true
# So what if you wanted to implement a different hashing function,
# or you've already built one and you want this pooler to use it?
#
# Current options:
#
# pg_bigint_hash: PARTITION BY HASH (Postgres hashing function)
# sha1: A hashing function based on SHA1
#
sharding_function = "pg_bigint_hash"
# Credentials for users that may connect to this cluster
[pools.sharded.users.0]
username = "postgres"
password = "postgres"
# Maximum number of server connections that can be established for this user
# The maximum number of connection from a single Pgcat process to any database in the cluster
# is the sum of pool_size across all users.
pool_size = 9
# Maximum query duration. Dangerous, but protects against DBs that died in a non-obvious way.
statement_timeout = 0
[pools.sharded.users.1]
username = "postgres"
password = "postgres"
pool_size = 21
statement_timeout = 15000
# Shard 0
[pools.sharded.shards.0]
# [ host, port, role ]
servers = [
[ "postgres", 5432, "primary" ],
[ "postgres", 5432, "replica" ]
]
# Database name (e.g. "postgres")
database = "postgres"
[pools.sharded.shards.1]
servers = [
[ "postgres", 5432, "primary" ],
[ "postgres", 5432, "replica" ],
]
database = "postgres"
[pools.sharded.shards.2]
servers = [
[ "postgres", 5432, "primary" ],
[ "postgres", 5432, "replica" ],
]
database = "postgres"
[pools.simple_db]
pool_mode = "session"
default_role = "primary"
query_parser_enabled = true
primary_reads_enabled = true
sharding_function = "pg_bigint_hash"
[pools.simple_db.users.0]
username = "postgres"
password = "postgres"
pool_size = 5
statement_timeout = 0
[pools.simple_db.shards.0]
servers = [
[ "postgres", 5432, "primary" ],
[ "postgres", 5432, "replica" ]
]
database = "postgres"

View File

@@ -5,55 +5,78 @@
#
# General pooler settings
[general]
# What IP to run on, 0.0.0.0 means accessible from everywhere.
host = "0.0.0.0"
# Port to run on, same as PgBouncer used in this example.
port = 6432
# Whether to enable prometheus exporter or not.
enable_prometheus_exporter = true
# How many connections to allocate per server.
pool_size = 15
# Port at which prometheus exporter listens on.
prometheus_exporter_port = 9930
# How long to wait before aborting a server connection (ms).
connect_timeout = 5000
# How much time to give the health check query to return with a result (ms).
healthcheck_timeout = 1000
# How long to keep connection available for immediate re-use, without running a healthcheck query on it
healthcheck_delay = 30000
# How much time to give clients during shutdown before forcibly killing client connections (ms).
shutdown_timeout = 60000
# For how long to ban a server if it fails a health check (seconds).
ban_time = 60 # seconds
# Reload config automatically if it changes.
autoreload = false
# TLS
# tls_certificate = "server.cert"
# tls_private_key = "server.key"
# Credentials to access the virtual administrative database (pgbouncer or pgcat)
# Connecting to that database allows running commands like `SHOW POOLS`, `SHOW DATABASES`, etc..
admin_username = "admin_user"
admin_password = "admin_pass"
# pool
# configs are structured as pool.<pool_name>
# the pool_name is what clients use as database name when connecting
# For the example below a client can connect using "postgres://sharding_user:sharding_user@pgcat_host:pgcat_port/sharded_db"
[pools.sharded_db]
# Pool mode (see PgBouncer docs for more).
# session: one server connection per connected client
# transaction: one server connection per client transaction
pool_mode = "transaction"
# How long to wait before aborting a server connection (ms).
connect_timeout = 5000
# How much time to give `SELECT 1` health check query to return with a result (ms).
healthcheck_timeout = 1000
# For how long to ban a server if it fails a health check (seconds).
ban_time = 60 # Seconds
# Stats will be sent here
statsd_address = "127.0.0.1:8125"
#
# User to use for authentication against the server.
[user]
name = "sharding_user"
password = "sharding_user"
#
# Shards in the cluster
[shards]
# Shard 0
[shards.0]
# [ host, port, role ]
servers = [
["127.0.0.1", 5432, "primary"],
["localhost", 5432, "replica"],
# [ "127.0.1.1", 5432, "replica" ],
]
# Database name (e.g. "postgres")
database = "shard0"
[shards.1]
# [ host, port, role ]
servers = [
["127.0.0.1", 5432, "primary"],
["localhost", 5432, "replica"],
# [ "127.0.1.1", 5432, "replica" ],
]
database = "shard1"
[shards.2]
# [ host, port, role ]
servers = [
["127.0.0.1", 5432, "primary"],
["localhost", 5432, "replica"],
# [ "127.0.1.1", 5432, "replica" ],
]
database = "shard2"
# Settings for our query routing layer.
[query_router]
# If the client doesn't specify, route traffic to
# this role by default.
#
@@ -62,15 +85,16 @@ pool_mode = "transaction"
# primary: all queries go to the primary unless otherwise specified.
default_role = "any"
# Query parser. If enabled, we'll attempt to parse
# every incoming query to determine if it's a read or a write.
# If it's a read query, we'll direct it to a replica. Otherwise, if it's a write,
# we'll direct it to the primary.
query_parser_enabled = true
query_parser_enabled = false
# If the query parser is enabled and this setting is enabled, the primary will be part of the pool of databases used for
# load balancing of read queries. Otherwise, the primary will only be used for write
# queries. The primary can always be explicitly selected with our custom protocol.
# queries. The primary can always be explicitely selected with our custom protocol.
primary_reads_enabled = true
# So what if you wanted to implement a different hashing function,
@@ -82,66 +106,3 @@ primary_reads_enabled = true
# sha1: A hashing function based on SHA1
#
sharding_function = "pg_bigint_hash"
# Credentials for users that may connect to this cluster
[pools.sharded_db.users.0]
username = "sharding_user"
password = "sharding_user"
# Maximum number of server connections that can be established for this user
# The maximum number of connection from a single Pgcat process to any database in the cluster
# is the sum of pool_size across all users.
pool_size = 9
# Maximum query duration. Dangerous, but protects against DBs that died in a non-obvious way.
statement_timeout = 0
[pools.sharded_db.users.1]
username = "other_user"
password = "other_user"
pool_size = 21
statement_timeout = 15000
# Shard 0
[pools.sharded_db.shards.0]
# [ host, port, role ]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ]
]
# Database name (e.g. "postgres")
database = "shard0"
[pools.sharded_db.shards.1]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ],
]
database = "shard1"
[pools.sharded_db.shards.2]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ],
]
database = "shard2"
[pools.simple_db]
pool_mode = "session"
default_role = "primary"
query_parser_enabled = true
primary_reads_enabled = true
sharding_function = "pg_bigint_hash"
[pools.simple_db.users.0]
username = "simple_user"
password = "simple_user"
pool_size = 5
statement_timeout = 0
[pools.simple_db.shards.0]
servers = [
[ "127.0.0.1", 5432, "primary" ],
[ "localhost", 5432, "replica" ]
]
database = "some_db"

View File

@@ -1,36 +1,21 @@
/// Admin database.
use bytes::{Buf, BufMut, BytesMut};
use log::{info, trace};
use tokio::net::tcp::OwnedWriteHalf;
use std::collections::HashMap;
use crate::config::{get_config, reload_config, VERSION};
use crate::config::{get_config, parse};
use crate::errors::Error;
use crate::messages::*;
use crate::pool::get_all_pools;
use crate::pool::ConnectionPool;
use crate::stats::get_stats;
use crate::ClientServerMap;
pub fn generate_server_info_for_admin() -> BytesMut {
let mut server_info = BytesMut::new();
server_info.put(server_paramater_message("application_name", ""));
server_info.put(server_paramater_message("client_encoding", "UTF8"));
server_info.put(server_paramater_message("server_encoding", "UTF8"));
server_info.put(server_paramater_message("server_version", VERSION));
server_info.put(server_paramater_message("DateStyle", "ISO, MDY"));
return server_info;
}
/// Handle admin client.
pub async fn handle_admin<T>(
stream: &mut T,
/// Handle admin client
pub async fn handle_admin(
stream: &mut OwnedWriteHalf,
mut query: BytesMut,
client_server_map: ClientServerMap,
) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
pool: ConnectionPool,
) -> Result<(), Error> {
let code = query.get_u8() as char;
if code != 'Q' {
@@ -44,86 +29,59 @@ where
trace!("Admin query: {}", query);
let query_parts: Vec<&str> = query.trim_end_matches(';').split_whitespace().collect();
match query_parts[0] {
"RELOAD" => {
trace!("RELOAD");
reload(stream, client_server_map).await
}
"SET" => {
trace!("SET");
ignore_set(stream).await
}
"SHOW" => match query_parts[1] {
"CONFIG" => {
trace!("SHOW CONFIG");
show_config(stream).await
}
"DATABASES" => {
trace!("SHOW DATABASES");
show_databases(stream).await
}
"LISTS" => {
trace!("SHOW LISTS");
show_lists(stream).await
}
"POOLS" => {
trace!("SHOW POOLS");
show_pools(stream).await
}
"STATS" => {
trace!("SHOW STATS");
show_stats(stream).await
}
"VERSION" => {
trace!("SHOW VERSION");
show_version(stream).await
}
_ => error_response(stream, "Unsupported SHOW query against the admin database").await,
},
_ => error_response(stream, "Unsupported query against the admin database").await,
if query.starts_with("SHOW STATS") {
trace!("SHOW STATS");
show_stats(stream).await
} else if query.starts_with("RELOAD") {
trace!("RELOAD");
reload(stream).await
} else if query.starts_with("SHOW CONFIG") {
trace!("SHOW CONFIG");
show_config(stream).await
} else if query.starts_with("SHOW DATABASES") {
trace!("SHOW DATABASES");
show_databases(stream, &pool).await
} else if query.starts_with("SHOW POOLS") {
trace!("SHOW POOLS");
show_pools(stream, &pool).await
} else if query.starts_with("SHOW LISTS") {
trace!("SHOW LISTS");
show_lists(stream, &pool).await
} else if query.starts_with("SHOW VERSION") {
trace!("SHOW VERSION");
show_version(stream).await
} else if query.starts_with("SET ") {
trace!("SET");
ignore_set(stream).await
} else {
error_response(stream, "Unsupported query against the admin database").await
}
}
/// Column-oriented statistics.
async fn show_lists<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// SHOW LISTS
async fn show_lists(stream: &mut OwnedWriteHalf, pool: &ConnectionPool) -> Result<(), Error> {
let stats = get_stats();
let columns = vec![("list", DataType::Text), ("items", DataType::Int4)];
let mut users = 1;
let mut databases = 1;
for (_, pool) in get_all_pools() {
databases += pool.databases();
users += 1; // One user per pool
}
let mut res = BytesMut::new();
res.put(row_description(&columns));
res.put(data_row(&vec![
"databases".to_string(),
databases.to_string(),
(pool.databases() + 1).to_string(), // see comment below
]));
res.put(data_row(&vec!["users".to_string(), users.to_string()]));
res.put(data_row(&vec!["pools".to_string(), databases.to_string()]));
res.put(data_row(&vec!["users".to_string(), "1".to_string()]));
res.put(data_row(&vec![
"pools".to_string(),
(pool.databases() + 1).to_string(), // +1 for the pgbouncer admin db pool which isn't real
])); // but admin tools that work with pgbouncer want this
res.put(data_row(&vec![
"free_clients".to_string(),
stats
.keys()
.map(|address_id| stats[&address_id]["cl_idle"])
.sum::<i64>()
.to_string(),
stats["cl_idle"].to_string(),
]));
res.put(data_row(&vec![
"used_clients".to_string(),
stats
.keys()
.map(|address_id| stats[&address_id]["cl_active"])
.sum::<i64>()
.to_string(),
stats["cl_active"].to_string(),
]));
res.put(data_row(&vec![
"login_clients".to_string(),
@@ -131,19 +89,11 @@ where
]));
res.put(data_row(&vec![
"free_servers".to_string(),
stats
.keys()
.map(|address_id| stats[&address_id]["sv_idle"])
.sum::<i64>()
.to_string(),
stats["sv_idle"].to_string(),
]));
res.put(data_row(&vec![
"used_servers".to_string(),
stats
.keys()
.map(|address_id| stats[&address_id]["sv_active"])
.sum::<i64>()
.to_string(),
stats["sv_active"].to_string(),
]));
res.put(data_row(&vec!["dns_names".to_string(), "0".to_string()]));
res.put(data_row(&vec!["dns_zones".to_string(), "0".to_string()]));
@@ -159,15 +109,12 @@ where
write_all_half(stream, res).await
}
/// Show PgCat version.
async fn show_version<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// SHOW VERSION
async fn show_version(stream: &mut OwnedWriteHalf) -> Result<(), Error> {
let mut res = BytesMut::new();
res.put(row_description(&vec![("version", DataType::Text)]));
res.put(data_row(&vec![format!("PgCat {}", VERSION).to_string()]));
res.put(data_row(&vec!["PgCat 0.1.0".to_string()]));
res.put(command_complete("SHOW"));
res.put_u8(b'Z');
@@ -177,17 +124,17 @@ where
write_all_half(stream, res).await
}
/// Show utilization of connection pools for each shard and replicas.
async fn show_pools<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// SHOW POOLS
async fn show_pools(stream: &mut OwnedWriteHalf, _pool: &ConnectionPool) -> Result<(), Error> {
let stats = get_stats();
let config = {
let guard = get_config();
&*guard.clone()
};
let columns = vec![
("database", DataType::Text),
("user", DataType::Text),
("cl_idle", DataType::Numeric),
("cl_active", DataType::Numeric),
("cl_waiting", DataType::Numeric),
("cl_cancel_req", DataType::Numeric),
@@ -203,32 +150,19 @@ where
let mut res = BytesMut::new();
res.put(row_description(&columns));
for (_, pool) in get_all_pools() {
let pool_config = &pool.settings;
for shard in 0..pool.shards() {
for server in 0..pool.servers(shard) {
let address = pool.address(shard, server);
let stats = match stats.get(&address.id) {
Some(stats) => stats.clone(),
None => HashMap::new(),
};
let mut row = vec![address.name(), pool_config.user.username.clone()];
let mut row = vec![String::from("all"), config.user.name.clone()];
for column in &columns[2..columns.len() - 1] {
let value = stats.get(column.0).unwrap_or(&0).to_string();
row.push(value);
}
row.push(pool_config.pool_mode.to_string());
res.put(data_row(&row));
}
}
for column in &columns[2..columns.len() - 1] {
let value = stats.get(column.0).unwrap_or(&0).to_string();
row.push(value);
}
row.push(config.general.pool_mode.to_string());
res.put(data_row(&row));
res.put(command_complete("SHOW"));
// ReadyForQuery
res.put_u8(b'Z');
res.put_i32(5);
res.put_u8(b'I');
@@ -236,11 +170,12 @@ where
write_all_half(stream, res).await
}
/// Show shards and replicas.
async fn show_databases<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// SHOW DATABASES
async fn show_databases(stream: &mut OwnedWriteHalf, pool: &ConnectionPool) -> Result<(), Error> {
let guard = get_config();
let config = &*guard.clone();
drop(guard);
// Columns
let columns = vec![
("name", DataType::Text),
@@ -260,39 +195,34 @@ where
let mut res = BytesMut::new();
// RowDescription
res.put(row_description(&columns));
for (_, pool) in get_all_pools() {
let pool_config = pool.settings.clone();
for shard in 0..pool.shards() {
let database_name = &pool.address(shard, 0).database;
for server in 0..pool.servers(shard) {
let address = pool.address(shard, server);
let pool_state = pool.pool_state(shard, server);
let banned = pool.is_banned(address, Some(address.role));
for shard in 0..pool.shards() {
let database_name = &config.shards[&shard.to_string()].database;
res.put(data_row(&vec![
address.name(), // name
address.host.to_string(), // host
address.port.to_string(), // port
database_name.to_string(), // database
pool_config.user.username.to_string(), // force_user
pool_config.user.pool_size.to_string(), // pool_size
"0".to_string(), // min_pool_size
"0".to_string(), // reserve_pool
pool_config.pool_mode.to_string(), // pool_mode
pool_config.user.pool_size.to_string(), // max_connections
pool_state.connections.to_string(), // current_connections
"0".to_string(), // paused
match banned {
// disabled
true => "1".to_string(),
false => "0".to_string(),
},
]));
}
for server in 0..pool.servers(shard) {
let address = pool.address(shard, server);
let pool_state = pool.pool_state(shard, server);
res.put(data_row(&vec![
address.name(), // name
address.host.to_string(), // host
address.port.to_string(), // port
database_name.to_string(), // database
config.user.name.to_string(), // force_user
config.general.pool_size.to_string(), // pool_size
"0".to_string(), // min_pool_size
"0".to_string(), // reserve_pool
config.general.pool_mode.to_string(), // pool_mode
config.general.pool_size.to_string(), // max_connections
pool_state.connections.to_string(), // current_connections
"0".to_string(), // paused
"0".to_string(), // disabled
]));
}
}
res.put(command_complete("SHOW"));
// ReadyForQuery
@@ -305,26 +235,26 @@ where
/// Ignore any SET commands the client sends.
/// This is common initialization done by ORMs.
async fn ignore_set<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
async fn ignore_set(stream: &mut OwnedWriteHalf) -> Result<(), Error> {
custom_protocol_response_ok(stream, "SET").await
}
/// Reload the configuration file without restarting the process.
async fn reload<T>(stream: &mut T, client_server_map: ClientServerMap) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// RELOAD
async fn reload(stream: &mut OwnedWriteHalf) -> Result<(), Error> {
info!("Reloading config");
reload_config(client_server_map).await?;
let config = get_config();
let path = config.path.clone().unwrap();
get_config().show();
parse(&path).await?;
let config = get_config();
config.show();
let mut res = BytesMut::new();
// CommandComplete
res.put(command_complete("RELOAD"));
// ReadyForQuery
@@ -335,15 +265,13 @@ where
write_all_half(stream, res).await
}
/// Shows current configuration.
async fn show_config<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
let config = &get_config();
async fn show_config(stream: &mut OwnedWriteHalf) -> Result<(), Error> {
let guard = get_config();
let config = &*guard.clone();
let config: HashMap<String, String> = config.into();
drop(guard);
// Configs that cannot be changed without restarting.
// Configs that cannot be changed dynamically.
let immutables = ["host", "port", "connect_timeout"];
// Columns
@@ -373,7 +301,6 @@ where
res.put(command_complete("SHOW"));
// ReadyForQuery
res.put_u8(b'Z');
res.put_i32(5);
res.put_u8(b'I');
@@ -381,14 +308,10 @@ where
write_all_half(stream, res).await
}
/// Show shard and replicas statistics.
async fn show_stats<T>(stream: &mut T) -> Result<(), Error>
where
T: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// SHOW STATS
async fn show_stats(stream: &mut OwnedWriteHalf) -> Result<(), Error> {
let columns = vec![
("database", DataType::Text),
("user", DataType::Text),
("total_xact_count", DataType::Numeric),
("total_query_count", DataType::Numeric),
("total_received", DataType::Numeric),
@@ -409,30 +332,17 @@ where
let mut res = BytesMut::new();
res.put(row_description(&columns));
for ((_db_name, username), pool) in get_all_pools() {
for shard in 0..pool.shards() {
for server in 0..pool.servers(shard) {
let address = pool.address(shard, server);
let stats = match stats.get(&address.id) {
Some(stats) => stats.clone(),
None => HashMap::new(),
};
let mut row = vec![
String::from("all"), // TODO: per-database stats,
];
let mut row = vec![address.name()];
row.push(username.clone());
for column in &columns[2..] {
row.push(stats.get(column.0).unwrap_or(&0).to_string());
}
res.put(data_row(&row));
}
}
for column in &columns[1..] {
row.push(stats.get(column.0).unwrap_or(&0).to_string());
}
res.put(data_row(&row));
res.put(command_complete("SHOW"));
// ReadyForQuery
res.put_u8(b'Z');
res.put_i32(5);
res.put_u8(b'I');

File diff suppressed because it is too large Load Diff

View File

@@ -1,27 +1,19 @@
/// Parse the configuration file.
use arc_swap::ArcSwap;
use arc_swap::{ArcSwap, Guard};
use log::{error, info};
use once_cell::sync::Lazy;
use serde_derive::{Deserialize, Serialize};
use std::collections::{HashMap, HashSet};
use std::hash::Hash;
use std::path::Path;
use std::sync::Arc;
use serde_derive::Deserialize;
use tokio::fs::File;
use tokio::io::AsyncReadExt;
use toml;
use std::collections::{HashMap, HashSet};
use std::sync::Arc;
use crate::errors::Error;
use crate::tls::{load_certs, load_keys};
use crate::{ClientServerMap, ConnectionPool};
pub const VERSION: &str = env!("CARGO_PKG_VERSION");
/// Globally available configuration.
static CONFIG: Lazy<ArcSwap<Config>> = Lazy::new(|| ArcSwap::from_pointee(Config::default()));
/// Server role: primary or replica.
#[derive(Clone, PartialEq, Serialize, Deserialize, Hash, std::cmp::Eq, Debug, Copy)]
#[derive(Clone, PartialEq, Deserialize, Hash, std::cmp::Eq, Debug, Copy)]
pub enum Role {
Primary,
Replica,
@@ -54,112 +46,62 @@ impl PartialEq<Role> for Option<Role> {
}
}
/// Address identifying a PostgreSQL server uniquely.
#[derive(Clone, PartialEq, Hash, std::cmp::Eq, Debug)]
pub struct Address {
/// Unique ID per addressable Postgres server.
pub id: usize,
/// Server host.
pub host: String,
/// Server port.
pub port: u16,
/// Shard number of this Postgres server.
pub port: String,
pub shard: usize,
/// The name of the Postgres database.
pub database: String,
/// Default search_path.
pub search_path: Option<String>,
/// Server role: replica, primary.
pub role: Role,
/// If it's a replica, number it for reference and failover.
pub replica_number: usize,
/// Position of the server in the pool for failover.
pub address_index: usize,
/// The name of the user configured to use this pool.
pub username: String,
/// The name of this pool (i.e. database name visible to the client).
pub pool_name: String,
}
impl Default for Address {
fn default() -> Address {
Address {
id: 0,
host: String::from("127.0.0.1"),
port: 5432,
port: String::from("5432"),
shard: 0,
address_index: 0,
replica_number: 0,
database: String::from("database"),
search_path: None,
role: Role::Replica,
username: String::from("username"),
pool_name: String::from("pool_name"),
}
}
}
impl Address {
/// Address name (aka database) used in `SHOW STATS`, `SHOW DATABASES`, and `SHOW POOLS`.
pub fn name(&self) -> String {
match self.role {
Role::Primary => format!("{}_shard_{}_primary", self.pool_name, self.shard),
Role::Primary => format!("shard_{}_primary", self.shard),
Role::Replica => format!(
"{}_shard_{}_replica_{}",
self.pool_name, self.shard, self.replica_number
),
Role::Replica => format!("shard_{}_replica_{}", self.shard, self.replica_number),
}
}
}
/// PostgreSQL user.
#[derive(Clone, PartialEq, Hash, std::cmp::Eq, Serialize, Deserialize, Debug)]
#[derive(Clone, PartialEq, Hash, std::cmp::Eq, Deserialize, Debug)]
pub struct User {
pub username: String,
pub name: String,
pub password: String,
pub pool_size: u32,
pub statement_timeout: u64,
}
impl Default for User {
fn default() -> User {
User {
username: String::from("postgres"),
name: String::from("postgres"),
password: String::new(),
pool_size: 15,
statement_timeout: 0,
}
}
}
/// General configuration.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
#[derive(Deserialize, Debug, Clone)]
pub struct General {
pub host: String,
pub port: i16,
pub enable_prometheus_exporter: Option<bool>,
pub prometheus_exporter_port: i16,
pub pool_size: u32,
pub pool_mode: String,
pub connect_timeout: u64,
pub healthcheck_timeout: u64,
pub shutdown_timeout: u64,
pub healthcheck_delay: u64,
pub ban_time: i64,
pub autoreload: bool,
pub tls_certificate: Option<String>,
pub tls_private_key: Option<String>,
pub admin_username: String,
pub admin_password: String,
pub statsd_address: String,
}
impl Default for General {
@@ -167,37 +109,42 @@ impl Default for General {
General {
host: String::from("localhost"),
port: 5432,
enable_prometheus_exporter: Some(false),
prometheus_exporter_port: 9930,
pool_size: 15,
pool_mode: String::from("transaction"),
connect_timeout: 5000,
healthcheck_timeout: 1000,
shutdown_timeout: 60000,
healthcheck_delay: 30000,
ban_time: 60,
autoreload: false,
tls_certificate: None,
tls_private_key: None,
admin_username: String::from("admin"),
admin_password: String::from("admin"),
statsd_address: String::from("127.0.0.1:8125"),
}
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
pub struct Pool {
pub pool_mode: String,
#[derive(Deserialize, Debug, Clone)]
pub struct Shard {
pub servers: Vec<(String, u16, String)>,
pub database: String,
}
impl Default for Shard {
fn default() -> Shard {
Shard {
servers: vec![(String::from("localhost"), 5432, String::from("primary"))],
database: String::from("postgres"),
}
}
}
#[derive(Deserialize, Debug, Clone)]
pub struct QueryRouter {
pub default_role: String,
pub query_parser_enabled: bool,
pub primary_reads_enabled: bool,
pub sharding_function: String,
pub shards: HashMap<String, Shard>,
pub users: HashMap<String, User>,
}
impl Default for Pool {
fn default() -> Pool {
Pool {
pool_mode: String::from("transaction"),
shards: HashMap::from([(String::from("1"), Shard::default())]),
users: HashMap::default(),
impl Default for QueryRouter {
fn default() -> QueryRouter {
QueryRouter {
default_role: String::from("any"),
query_parser_enabled: false,
primary_reads_enabled: true,
@@ -206,109 +153,39 @@ impl Default for Pool {
}
}
/// Shard configuration.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
pub struct Shard {
pub database: String,
pub search_path: Option<String>,
pub servers: Vec<(String, u16, String)>,
}
impl Default for Shard {
fn default() -> Shard {
Shard {
servers: vec![(String::from("localhost"), 5432, String::from("primary"))],
search_path: None,
database: String::from("postgres"),
}
}
}
fn default_path() -> String {
String::from("pgcat.toml")
}
/// Configuration wrapper.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
#[derive(Deserialize, Debug, Clone)]
pub struct Config {
// Serializer maintains the order of fields in the struct
// so we should always put simple fields before nested fields
// in all serializable structs to avoid ValueAfterTable errors
// These errors occur when the toml serializer is about to produce
// ambigous toml structure like the one below
// [main]
// field1_under_main = 1
// field2_under_main = 2
// [main.subconf]
// field1_under_subconf = 1
// field3_under_main = 3 # This field will be interpreted as being under subconf and not under main
#[serde(default = "default_path")]
pub path: String,
pub path: Option<String>,
pub general: General,
pub pools: HashMap<String, Pool>,
pub user: User,
pub shards: HashMap<String, Shard>,
pub query_router: QueryRouter,
}
impl Default for Config {
fn default() -> Config {
Config {
path: String::from("pgcat.toml"),
path: Some(String::from("pgcat.toml")),
general: General::default(),
pools: HashMap::default(),
user: User::default(),
shards: HashMap::from([(String::from("1"), Shard::default())]),
query_router: QueryRouter::default(),
}
}
}
impl From<&Config> for std::collections::HashMap<String, String> {
fn from(config: &Config) -> HashMap<String, String> {
let mut r: Vec<(String, String)> = config
.pools
.iter()
.flat_map(|(pool_name, pool)| {
[
(
format!("pools.{}.pool_mode", pool_name),
pool.pool_mode.clone(),
),
(
format!("pools.{}.primary_reads_enabled", pool_name),
pool.primary_reads_enabled.to_string(),
),
(
format!("pools.{}.query_parser_enabled", pool_name),
pool.query_parser_enabled.to_string(),
),
(
format!("pools.{}.default_role", pool_name),
pool.default_role.clone(),
),
(
format!("pools.{}.sharding_function", pool_name),
pool.sharding_function.clone(),
),
(
format!("pools.{:?}.shard_count", pool_name),
pool.shards.len().to_string(),
),
(
format!("pools.{:?}.users", pool_name),
pool.users
.iter()
.map(|(_username, user)| &user.username)
.cloned()
.collect::<Vec<String>>()
.join(", "),
),
]
})
.collect();
let mut static_settings = vec![
HashMap::from([
("host".to_string(), config.general.host.to_string()),
("port".to_string(), config.general.port.to_string()),
(
"prometheus_exporter_port".to_string(),
config.general.prometheus_exporter_port.to_string(),
"pool_size".to_string(),
config.general.pool_size.to_string(),
),
(
"pool_mode".to_string(),
config.general.pool_mode.to_string(),
),
(
"connect_timeout".to_string(),
@@ -318,110 +195,51 @@ impl From<&Config> for std::collections::HashMap<String, String> {
"healthcheck_timeout".to_string(),
config.general.healthcheck_timeout.to_string(),
),
(
"shutdown_timeout".to_string(),
config.general.shutdown_timeout.to_string(),
),
(
"healthcheck_delay".to_string(),
config.general.healthcheck_delay.to_string(),
),
("ban_time".to_string(), config.general.ban_time.to_string()),
];
r.append(&mut static_settings);
return r.iter().cloned().collect();
(
"statsd_address".to_string(),
config.general.statsd_address.to_string(),
),
(
"default_role".to_string(),
config.query_router.default_role.to_string(),
),
(
"query_parser_enabled".to_string(),
config.query_router.query_parser_enabled.to_string(),
),
(
"primary_reads_enabled".to_string(),
config.query_router.primary_reads_enabled.to_string(),
),
(
"sharding_function".to_string(),
config.query_router.sharding_function.to_string(),
),
])
}
}
impl Config {
/// Print current configuration.
pub fn show(&self) {
info!("Pool size: {}", self.general.pool_size);
info!("Pool mode: {}", self.general.pool_mode);
info!("Ban time: {}s", self.general.ban_time);
info!(
"Healthcheck timeout: {}ms",
self.general.healthcheck_timeout
);
info!("Connection timeout: {}ms", self.general.connect_timeout);
info!("Shutdown timeout: {}ms", self.general.shutdown_timeout);
info!("Healthcheck delay: {}ms", self.general.healthcheck_delay);
match self.general.tls_certificate.clone() {
Some(tls_certificate) => {
info!("TLS certificate: {}", tls_certificate);
match self.general.tls_private_key.clone() {
Some(tls_private_key) => {
info!("TLS private key: {}", tls_private_key);
info!("TLS support is enabled");
}
None => (),
}
}
None => {
info!("TLS support is disabled");
}
};
for (pool_name, pool_config) in &self.pools {
// TODO: Make this output prettier (maybe a table?)
info!(
"[pool: {}] Maximum user connections: {}",
pool_name,
pool_config
.users
.iter()
.map(|(_, user_cfg)| user_cfg.pool_size)
.sum::<u32>()
.to_string()
);
info!("[pool: {}] Pool mode: {}", pool_name, pool_config.pool_mode);
info!(
"[pool: {}] Sharding function: {}",
pool_name, pool_config.sharding_function
);
info!(
"[pool: {}] Primary reads: {}",
pool_name, pool_config.primary_reads_enabled
);
info!(
"[pool: {}] Query router: {}",
pool_name, pool_config.query_parser_enabled
);
info!(
"[pool: {}] Number of shards: {}",
pool_name,
pool_config.shards.len()
);
info!(
"[pool: {}] Number of users: {}",
pool_name,
pool_config.users.len()
);
for user in &pool_config.users {
info!(
"[pool: {}][user: {}] Pool size: {}",
pool_name, user.1.username, user.1.pool_size,
);
info!(
"[pool: {}][user: {}] Statement timeout: {}",
pool_name, user.1.username, user.1.statement_timeout
)
}
}
info!("Sharding function: {}", self.query_router.sharding_function);
info!("Number of shards: {}", self.shards.len());
}
}
/// Get a read-only instance of the configuration
/// from anywhere in the app.
/// ArcSwap makes this cheap and quick.
pub fn get_config() -> Config {
(*(*CONFIG.load())).clone()
pub fn get_config() -> Guard<Arc<Config>> {
CONFIG.load()
}
/// Parse the configuration file located at the path.
/// Parse the config.
pub async fn parse(path: &str) -> Result<(), Error> {
let mut contents = String::new();
let mut file = match File::open(path).await {
@@ -448,161 +266,93 @@ pub async fn parse(path: &str) -> Result<(), Error> {
}
};
// Validate TLS!
match config.general.tls_certificate.clone() {
Some(tls_certificate) => {
match load_certs(&Path::new(&tls_certificate)) {
Ok(_) => {
// Cert is okay, but what about the private key?
match config.general.tls_private_key.clone() {
Some(tls_private_key) => match load_keys(&Path::new(&tls_private_key)) {
Ok(_) => (),
Err(err) => {
error!("tls_private_key is incorrectly configured: {:?}", err);
return Err(Error::BadConfig);
}
},
None => {
error!("tls_certificate is set, but the tls_private_key is not");
return Err(Error::BadConfig);
}
};
}
Err(err) => {
error!("tls_certificate is incorrectly configured: {:?}", err);
return Err(Error::BadConfig);
}
}
match config.query_router.sharding_function.as_ref() {
"pg_bigint_hash" => (),
"sha1" => (),
_ => {
error!(
"Supported sharding functions are: 'pg_bigint_hash', 'sha1', got: '{}'",
config.query_router.sharding_function
);
return Err(Error::BadConfig);
}
None => (),
};
for (pool_name, pool) in &config.pools {
match pool.sharding_function.as_ref() {
"pg_bigint_hash" => (),
"sha1" => (),
_ => {
// Quick config sanity check.
for shard in &config.shards {
// We use addresses as unique identifiers,
// let's make sure they are unique in the config as well.
let mut dup_check = HashSet::new();
let mut primary_count = 0;
match shard.0.parse::<usize>() {
Ok(_) => (),
Err(_) => {
error!(
"Supported sharding functions are: 'pg_bigint_hash', 'sha1', got: '{}' in pool {} settings",
pool.sharding_function,
pool_name
"Shard '{}' is not a valid number, shards must be numbered starting at 0",
shard.0
);
return Err(Error::BadConfig);
}
};
match pool.default_role.as_ref() {
"any" => (),
"primary" => (),
"replica" => (),
other => {
error!(
"Query router default_role must be 'primary', 'replica', or 'any', got: '{}'",
other
);
return Err(Error::BadConfig);
}
};
if shard.1.servers.len() == 0 {
error!("Shard {} has no servers configured", shard.0);
return Err(Error::BadConfig);
}
match pool.pool_mode.as_ref() {
"transaction" => (),
"session" => (),
other => {
error!(
"pool_mode can be 'session' or 'transaction', got: '{}'",
other
);
return Err(Error::BadConfig);
}
};
for server in &shard.1.servers {
dup_check.insert(server);
for shard in &pool.shards {
// We use addresses as unique identifiers,
// let's make sure they are unique in the config as well.
let mut dup_check = HashSet::new();
let mut primary_count = 0;
// Check that we define only zero or one primary.
match server.2.as_ref() {
"primary" => primary_count += 1,
_ => (),
};
match shard.0.parse::<usize>() {
Ok(_) => (),
Err(_) => {
// Check role spelling.
match server.2.as_ref() {
"primary" => (),
"replica" => (),
_ => {
error!(
"Shard '{}' is not a valid number, shards must be numbered starting at 0",
shard.0
"Shard {} server role must be either 'primary' or 'replica', got: '{}'",
shard.0, server.2
);
return Err(Error::BadConfig);
}
};
}
if shard.1.servers.len() == 0 {
error!("Shard {} has no servers configured", shard.0);
return Err(Error::BadConfig);
}
if primary_count > 1 {
error!("Shard {} has more than on primary configured", &shard.0);
return Err(Error::BadConfig);
}
for server in &shard.1.servers {
dup_check.insert(server);
// Check that we define only zero or one primary.
match server.2.as_ref() {
"primary" => primary_count += 1,
_ => (),
};
// Check role spelling.
match server.2.as_ref() {
"primary" => (),
"replica" => (),
_ => {
error!(
"Shard {} server role must be either 'primary' or 'replica', got: '{}'",
shard.0, server.2
);
return Err(Error::BadConfig);
}
};
}
if primary_count > 1 {
error!("Shard {} has more than on primary configured", &shard.0);
return Err(Error::BadConfig);
}
if dup_check.len() != shard.1.servers.len() {
error!("Shard {} contains duplicate server configs", &shard.0);
return Err(Error::BadConfig);
}
if dup_check.len() != shard.1.servers.len() {
error!("Shard {} contains duplicate server configs", &shard.0);
return Err(Error::BadConfig);
}
}
config.path = path.to_string();
// Update the configuration globally.
CONFIG.store(Arc::new(config.clone()));
Ok(())
}
pub async fn reload_config(client_server_map: ClientServerMap) -> Result<bool, Error> {
let old_config = get_config();
match parse(&old_config.path).await {
Ok(()) => (),
Err(err) => {
error!("Config reload error: {:?}", err);
match config.query_router.default_role.as_ref() {
"any" => (),
"primary" => (),
"replica" => (),
other => {
error!(
"Query router default_role must be 'primary', 'replica', or 'any', got: '{}'",
other
);
return Err(Error::BadConfig);
}
};
let new_config = get_config();
if old_config.pools != new_config.pools {
info!("Pool configuration changed, re-creating server pools");
ConnectionPool::from_config(client_server_map).await?;
Ok(true)
} else if old_config != new_config {
Ok(true)
} else {
Ok(false)
}
config.path = Some(path.to_string());
CONFIG.store(Arc::new(config.clone()));
Ok(())
}
#[cfg(test)]
@@ -612,67 +362,11 @@ mod test {
#[tokio::test]
async fn test_config() {
parse("pgcat.toml").await.unwrap();
assert_eq!(get_config().path, "pgcat.toml".to_string());
assert_eq!(get_config().general.ban_time, 60);
assert_eq!(get_config().pools.len(), 2);
assert_eq!(get_config().pools["sharded_db"].shards.len(), 3);
assert_eq!(get_config().pools["simple_db"].shards.len(), 1);
assert_eq!(get_config().pools["sharded_db"].users.len(), 2);
assert_eq!(get_config().pools["simple_db"].users.len(), 1);
assert_eq!(
get_config().pools["sharded_db"].shards["0"].servers[0].0,
"127.0.0.1"
);
assert_eq!(
get_config().pools["sharded_db"].shards["1"].servers[0].2,
"primary"
);
assert_eq!(
get_config().pools["sharded_db"].shards["1"].database,
"shard1"
);
assert_eq!(
get_config().pools["sharded_db"].users["0"].username,
"sharding_user"
);
assert_eq!(
get_config().pools["sharded_db"].users["1"].password,
"other_user"
);
assert_eq!(get_config().pools["sharded_db"].users["1"].pool_size, 21);
assert_eq!(get_config().pools["sharded_db"].default_role, "any");
assert_eq!(
get_config().pools["simple_db"].shards["0"].servers[0].0,
"127.0.0.1"
);
assert_eq!(
get_config().pools["simple_db"].shards["0"].servers[0].1,
5432
);
assert_eq!(
get_config().pools["simple_db"].shards["0"].database,
"some_db"
);
assert_eq!(get_config().pools["simple_db"].default_role, "primary");
assert_eq!(
get_config().pools["simple_db"].users["0"].username,
"simple_user"
);
assert_eq!(
get_config().pools["simple_db"].users["0"].password,
"simple_user"
);
assert_eq!(get_config().pools["simple_db"].users["0"].pool_size, 5);
}
#[tokio::test]
async fn test_serialize_configs() {
parse("pgcat.toml").await.unwrap();
print!("{}", toml::to_string(&get_config()).unwrap());
assert_eq!(get_config().general.pool_size, 15);
assert_eq!(get_config().shards.len(), 3);
assert_eq!(get_config().shards["1"].servers[0].0, "127.0.0.1");
assert_eq!(get_config().shards["0"].servers[0].2, "primary");
assert_eq!(get_config().query_router.default_role, "any");
assert_eq!(get_config().path, Some("pgcat.toml".to_string()));
}
}

View File

@@ -1,6 +1,7 @@
/// Various protocol constants, as defined in
/// <https://www.postgresql.org/docs/12/protocol-message-formats.html>
/// https://www.postgresql.org/docs/12/protocol-message-formats.html
/// and elsewhere in the source code.
/// Also other constants we use elsewhere.
// Used in the StartupMessage to indicate regular handshake.
pub const PROTOCOL_VERSION_NUMBER: i32 = 196608;
@@ -14,13 +15,6 @@ pub const CANCEL_REQUEST_CODE: i32 = 80877102;
// AuthenticationMD5Password
pub const MD5_ENCRYPTED_PASSWORD: i32 = 5;
// SASL
pub const SASL: i32 = 10;
pub const SASL_CONTINUE: i32 = 11;
pub const SASL_FINAL: i32 = 12;
pub const SCRAM_SHA_256: &str = "SCRAM-SHA-256";
pub const NONCE_LENGTH: usize = 24;
// AuthenticationOk
pub const AUTHENTICATION_SUCCESSFUL: i32 = 0;

View File

@@ -1,16 +1,14 @@
/// Errors.
/// Various errors.
#[derive(Debug, PartialEq)]
pub enum Error {
SocketError,
// ClientDisconnected,
ClientBadStartup,
ProtocolSyncError,
ServerError,
// ServerTimeout,
// DirtyServer,
BadConfig,
BadUserList,
AllServersDown,
ClientError,
TlsError,
StatementTimeout,
ShuttingDown,
AuthenticationError
}

View File

@@ -1,4 +1,4 @@
// Copyright (c) 2022 Lev Kokotov <hi@levthe.dev>
// Copyright (c) 2022 Lev Kokotov <lev@levthe.dev>
// Permission is hereby granted, free of charge, to any person obtaining
// a copy of this software and associated documentation files (the
@@ -24,62 +24,58 @@ extern crate async_trait;
extern crate bb8;
extern crate bytes;
extern crate env_logger;
extern crate exitcode;
extern crate log;
extern crate md5;
extern crate num_cpus;
extern crate once_cell;
extern crate rustls_pemfile;
extern crate serde;
extern crate serde_derive;
extern crate sqlparser;
extern crate statsd;
extern crate tokio;
extern crate tokio_rustls;
extern crate toml;
use log::{debug, error, info};
use log::{error, info};
use parking_lot::Mutex;
use tokio::net::TcpListener;
use tokio::{
signal,
signal::unix::{signal as unix_signal, SignalKind},
sync::mpsc,
};
use std::collections::HashMap;
use std::net::SocketAddr;
use std::str::FromStr;
use std::sync::Arc;
use tokio::sync::broadcast;
mod admin;
mod client;
mod config;
mod userlist;
mod constants;
mod errors;
mod messages;
mod pool;
mod prometheus;
mod query_router;
mod scram;
mod server;
mod sharding;
mod stats;
mod tls;
use crate::config::{get_config, reload_config, VERSION};
use crate::errors::Error;
use crate::pool::{ClientServerMap, ConnectionPool};
use crate::prometheus::start_metric_server;
use crate::stats::{Collector, Reporter, REPORTER};
// Support for query cancellation: this maps our process_ids and
// secret keys to the backend's.
use config::get_config;
use pool::{ClientServerMap, ConnectionPool};
use stats::{Collector, Reporter};
/// Main!
#[tokio::main(worker_threads = 4)]
async fn main() {
env_logger::init();
info!("Welcome to PgCat! Meow. (Version {})", VERSION);
info!("Welcome to PgCat! Meow.");
// Prepare regexes
if !query_router::QueryRouter::setup() {
error!("Could not setup query router");
std::process::exit(exitcode::CONFIG);
return;
}
let args = std::env::args().collect::<Vec<String>>();
@@ -90,208 +86,142 @@ async fn main() {
String::from("pgcat.toml")
};
// Prepare the config
match config::parse(&config_file).await {
Ok(_) => (),
Err(err) => {
error!("Config parse error: {:?}", err);
std::process::exit(exitcode::CONFIG);
return;
}
};
// Prepare user list
match userlist::parse("userlist.json").await {
Ok(_) => (),
Err(err) => {
error!("Userlist parse error: {:?}", err);
return;
}
};
let config = get_config();
if let Some(true) = config.general.enable_prometheus_exporter {
let http_addr_str = format!(
"{}:{}",
config.general.host, config.general.prometheus_exporter_port
);
let http_addr = match SocketAddr::from_str(&http_addr_str) {
Ok(addr) => addr,
Err(err) => {
error!("Invalid http address: {}", err);
std::process::exit(exitcode::CONFIG);
}
};
tokio::task::spawn(async move {
start_metric_server(http_addr).await;
});
}
let addr = format!("{}:{}", config.general.host, config.general.port);
let listener = match TcpListener::bind(&addr).await {
Ok(sock) => sock,
Err(err) => {
error!("Listener socket error: {:?}", err);
std::process::exit(exitcode::CONFIG);
return;
}
};
info!("Running on {}", addr);
config.show();
// Tracks which client is connected to which server for query cancellation.
let client_server_map: ClientServerMap = Arc::new(Mutex::new(HashMap::new()));
// Statistics reporting.
let (stats_tx, stats_rx) = mpsc::channel(100_000);
REPORTER.store(Arc::new(Reporter::new(stats_tx.clone())));
// Connection pool that allows to query all shards and replicas.
match ConnectionPool::from_config(client_server_map.clone()).await {
Ok(_) => (),
Err(err) => {
error!("Pool error: {:?}", err);
std::process::exit(exitcode::CONFIG);
}
};
// Collect statistics and send them to StatsD
let (tx, rx) = mpsc::channel(100);
let collector_tx = tx.clone();
tokio::task::spawn(async move {
let mut stats_collector = Collector::new(stats_rx, stats_tx.clone());
let mut stats_collector = Collector::new(rx, collector_tx);
stats_collector.collect().await;
});
info!("Config autoreloader: {}", config.general.autoreload);
let mut pool =
ConnectionPool::from_config(client_server_map.clone(), Reporter::new(tx.clone())).await;
let mut term_signal = unix_signal(SignalKind::terminate()).unwrap();
let mut interrupt_signal = unix_signal(SignalKind::interrupt()).unwrap();
let mut sighup_signal = unix_signal(SignalKind::hangup()).unwrap();
let mut autoreload_interval = tokio::time::interval(tokio::time::Duration::from_millis(15_000));
let (shutdown_tx, _) = broadcast::channel::<()>(1);
let (drain_tx, mut drain_rx) = mpsc::channel::<i8>(2048);
let (exit_tx, mut exit_rx) = mpsc::channel::<()>(1);
let server_info = match pool.validate().await {
Ok(info) => info,
Err(err) => {
error!("Could not validate connection pool: {:?}", err);
return;
}
};
info!("Waiting for clients");
let mut admin_only = false;
let mut total_clients = 0;
// Main app runs here.
tokio::task::spawn(async move {
loop {
let pool = pool.clone();
let client_server_map = client_server_map.clone();
let server_info = server_info.clone();
let reporter = Reporter::new(tx.clone());
loop {
tokio::select! {
// Reload config:
// kill -SIGHUP $(pgrep pgcat)
_ = sighup_signal.recv() => {
info!("Reloading config");
let (socket, addr) = match listener.accept().await {
Ok((socket, addr)) => (socket, addr),
Err(err) => {
error!("{:?}", err);
continue;
}
};
match reload_config(client_server_map.clone()).await {
Ok(_) => (),
Err(_) => (),
};
// Client goes to another thread, bye.
tokio::task::spawn(async move {
let start = chrono::offset::Utc::now().naive_utc();
match client::Client::startup(socket, client_server_map, server_info, reporter)
.await
{
Ok(mut client) => {
info!("Client {:?} connected", addr);
match client.handle(pool).await {
Ok(()) => {
let duration = chrono::offset::Utc::now().naive_utc() - start;
get_config().show();
},
info!(
"Client {:?} disconnected, session duration: {}",
addr,
format_duration(&duration)
);
}
_ = autoreload_interval.tick() => {
if config.general.autoreload {
info!("Automatically reloading config");
match reload_config(client_server_map.clone()).await {
Ok(changed) => {
if changed {
get_config().show()
Err(err) => {
error!("Client disconnected with error: {:?}", err);
client.release();
}
}
Err(_) => (),
};
}
},
}
// Initiate graceful shutdown sequence on sig int
_ = interrupt_signal.recv() => {
info!("Got SIGINT, waiting for client connection drain now");
admin_only = true;
// Broadcast that client tasks need to finish
let _ = shutdown_tx.send(());
let exit_tx = exit_tx.clone();
let _ = drain_tx.send(0).await;
tokio::task::spawn(async move {
let mut interval = tokio::time::interval(tokio::time::Duration::from_millis(config.general.shutdown_timeout));
// First tick fires immediately.
interval.tick().await;
// Second one in the interval time.
interval.tick().await;
// We're done waiting.
error!("Timed out waiting for clients");
let _ = exit_tx.send(()).await;
});
},
_ = term_signal.recv() => break,
new_client = listener.accept() => {
let (socket, addr) = match new_client {
Ok((socket, addr)) => (socket, addr),
Err(err) => {
error!("{:?}", err);
continue;
error!("Client failed to login: {:?}", err);
}
};
let shutdown_rx = shutdown_tx.subscribe();
let drain_tx = drain_tx.clone();
let client_server_map = client_server_map.clone();
tokio::task::spawn(async move {
let start = chrono::offset::Utc::now().naive_utc();
match client::client_entrypoint(
socket,
client_server_map,
shutdown_rx,
drain_tx,
admin_only,
)
.await
{
Ok(()) => {
let duration = chrono::offset::Utc::now().naive_utc() - start;
info!(
"Client {:?} disconnected, session duration: {}",
addr,
format_duration(&duration)
);
}
Err(err) => {
match err {
// Don't count the clients we rejected.
Error::ShuttingDown => (),
_ => {
// drain_tx.send(-1).await.unwrap();
}
}
debug!("Client disconnected with error {:?}", err);
}
};
});
}
_ = exit_rx.recv() => {
break;
}
client_ping = drain_rx.recv() => {
let client_ping = client_ping.unwrap();
total_clients += client_ping;
if total_clients == 0 && admin_only {
let _ = exit_tx.send(()).await;
}
}
});
}
}
});
info!("Shutting down...");
// Reload config
// kill -SIGHUP $(pgrep pgcat)
tokio::task::spawn(async move {
let mut stream = unix_signal(SignalKind::hangup()).unwrap();
loop {
stream.recv().await;
info!("Reloading config");
match config::parse("pgcat.toml").await {
Ok(_) => {
get_config().show();
}
Err(err) => {
error!("{:?}", err);
return;
}
};
}
});
// Setup shut down sequence
match signal::ctrl_c().await {
Ok(()) => {
info!("Shutting down...");
}
Err(err) => {
error!("Unable to listen for shutdown signal: {}", err);
}
};
}
/// Format chrono::Duration to be more human-friendly.

View File

@@ -1,13 +1,23 @@
/// Helper functions to send one-off protocol messages
/// and handle TcpStream (TCP socket).
use bytes::{Buf, BufMut, BytesMut};
use md5::{Digest, Md5};
use tokio::io::{AsyncReadExt, AsyncWriteExt};
use tokio::net::TcpStream;
use tokio::io::{AsyncReadExt, AsyncWriteExt, BufReader};
use tokio::net::{
tcp::{OwnedReadHalf, OwnedWriteHalf},
TcpStream,
};
use log::{error};
use crate::errors::Error;
use std::collections::HashMap;
use std::mem;
use rand::Rng;
use crate::userlist::get_user_list;
/// Postgres data type mappings
/// used in RowDescription ('T') message.
@@ -27,11 +37,100 @@ impl From<&DataType> for i32 {
}
}
/**
1. Generate salt (4 bytes of random data)
md5(concat(md5(concat(password, username)), random-salt)))
2. Send md5 auth request
3. recieve PasswordMessage with salt.
4. refactor md5_password function to be reusable
5. check username hash combo against file
6. AuthenticationOk or ErrorResponse
**/
pub async fn start_auth(stream: &mut TcpStream, user_name: &String) -> Result<(), Error> {
let mut rng = rand::thread_rng();
//Generate random 4 byte salt
let salt = rng.gen::<u32>();
// Send AuthenticationMD5Password request
send_md5_request(stream, salt).await?;
let code = match stream.read_u8().await {
Ok(code) => code as char,
Err(_) => return Err(Error::AuthenticationError),
};
match code {
// Password response
'p' => {
fetch_password_and_authenticate(stream, &user_name, &salt).await?;
Ok(auth_ok(stream).await?)
}
_ => {
error!("Unknown code: {}", code);
return Err(Error::AuthenticationError);
}
}
}
pub async fn send_md5_request(stream: &mut TcpStream, salt: u32) -> Result<(), Error> {
let mut authentication_md5password = BytesMut::with_capacity(12);
authentication_md5password.put_u8(b'R');
authentication_md5password.put_i32(12);
authentication_md5password.put_i32(5);
authentication_md5password.put_u32(salt);
// Send AuthenticationMD5Password request
Ok(write_all(stream, authentication_md5password).await?)
}
pub async fn fetch_password_and_authenticate(stream: &mut TcpStream, user_name: &String, salt: &u32) -> Result<(), Error> {
/**
1. How do I store the lists of users and paswords? clear text or hash?? wtf
2. Add auth to tests
**/
let len = match stream.read_i32().await {
Ok(len) => len,
Err(_) => return Err(Error::AuthenticationError),
};
// Read whatever is left.
let mut password_hash = vec![0u8; len as usize - 4];
match stream.read_exact(&mut password_hash).await {
Ok(_) => (),
Err(_) => return Err(Error::AuthenticationError),
};
let user_list = get_user_list();
let mut password: String = String::new();
match user_list.get(&user_name) {
Some(&p) => password = p,
None => return Err(Error::AuthenticationError),
}
let mut md5 = Md5::new();
// concat('md5', md5(concat(md5(concat(password, username)), random-salt)))
// First pass
md5.update(&password.as_bytes());
md5.update(&user_name.as_bytes());
let output = md5.finalize_reset();
// Second pass
md5.update(format!("{:x}", output));
md5.update(salt.to_be_bytes().to_vec());
let password_string: String = String::from_utf8(password_hash).expect("Could not get password hash");
match format!("md5{:x}", md5.finalize()) == password_string {
true => Ok(()),
_ => Err(Error::AuthenticationError)
}
}
/// Tell the client that authentication handshake completed successfully.
pub async fn auth_ok<S>(stream: &mut S) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
pub async fn auth_ok(stream: &mut TcpStream) -> Result<(), Error> {
let mut auth_ok = BytesMut::with_capacity(9);
auth_ok.put_u8(b'R');
@@ -41,39 +140,13 @@ where
Ok(write_all(stream, auth_ok).await?)
}
/// Generate md5 password challenge.
pub async fn md5_challenge<S>(stream: &mut S) -> Result<[u8; 4], Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
// let mut rng = rand::thread_rng();
let salt: [u8; 4] = [
rand::random(),
rand::random(),
rand::random(),
rand::random(),
];
let mut res = BytesMut::new();
res.put_u8(b'R');
res.put_i32(12);
res.put_i32(5); // MD5
res.put_slice(&salt[..]);
write_all(stream, res).await?;
Ok(salt)
}
/// Give the client the process_id and secret we generated
/// used in query cancellation.
pub async fn backend_key_data<S>(
stream: &mut S,
pub async fn backend_key_data(
stream: &mut TcpStream,
backend_id: i32,
secret_key: i32,
) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
) -> Result<(), Error> {
let mut key_data = BytesMut::from(&b"K"[..]);
key_data.put_i32(12);
key_data.put_i32(backend_id);
@@ -94,13 +167,8 @@ pub fn simple_query(query: &str) -> BytesMut {
}
/// Tell the client we're ready for another query.
pub async fn ready_for_query<S>(stream: &mut S) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
let mut bytes = BytesMut::with_capacity(
mem::size_of::<u8>() + mem::size_of::<i32>() + mem::size_of::<u8>(),
);
pub async fn ready_for_query(stream: &mut TcpStream) -> Result<(), Error> {
let mut bytes = BytesMut::with_capacity(5);
bytes.put_u8(b'Z');
bytes.put_i32(5);
@@ -111,12 +179,7 @@ where
/// Send the startup packet the server. We're pretending we're a Pg client.
/// This tells the server which user we are and what database we want.
pub async fn startup(
stream: &mut TcpStream,
user: &str,
database: &str,
search_path: Option<&String>,
) -> Result<(), Error> {
pub async fn startup(stream: &mut TcpStream, user: &str, database: &str) -> Result<(), Error> {
let mut bytes = BytesMut::with_capacity(25);
bytes.put_i32(196608); // Protocol number
@@ -130,17 +193,6 @@ pub async fn startup(
bytes.put(&b"database\0"[..]);
bytes.put_slice(&database.as_bytes());
bytes.put_u8(0);
// search_path
match search_path {
Some(search_path) => {
bytes.put(&b"options\0"[..]);
bytes.put_slice(&format!("-c search_path={}", search_path).as_bytes());
bytes.put_u8(0);
}
None => (),
};
bytes.put_u8(0); // Null terminator
let len = bytes.len() as i32 + 4i32;
@@ -208,8 +260,14 @@ pub fn parse_startup(bytes: BytesMut) -> Result<HashMap<String, String>, Error>
Ok(result)
}
/// Create md5 password hash given a salt.
pub fn md5_hash_password(user: &str, password: &str, salt: &[u8]) -> Vec<u8> {
/// Send password challenge response to the server.
/// This is the MD5 challenge.
pub async fn md5_password(
stream: &mut TcpStream,
user: &str,
password: &str,
salt: &[u8],
) -> Result<(), Error> {
let mut md5 = Md5::new();
// First pass
@@ -228,22 +286,6 @@ pub fn md5_hash_password(user: &str, password: &str, salt: &[u8]) -> Vec<u8> {
.collect::<Vec<u8>>();
password.push(0);
password
}
/// Send password challenge response to the server.
/// This is the MD5 challenge.
pub async fn md5_password<S>(
stream: &mut S,
user: &str,
password: &str,
salt: &[u8],
) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
let password = md5_hash_password(user, password, salt);
let mut message = BytesMut::with_capacity(password.len() as usize + 5);
message.put_u8(b'p');
@@ -256,10 +298,10 @@ where
/// Implements a response to our custom `SET SHARDING KEY`
/// and `SET SERVER ROLE` commands.
/// This tells the client we're ready for the next query.
pub async fn custom_protocol_response_ok<S>(stream: &mut S, message: &str) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
pub async fn custom_protocol_response_ok(
stream: &mut OwnedWriteHalf,
message: &str,
) -> Result<(), Error> {
let mut res = BytesMut::with_capacity(25);
let set_complete = BytesMut::from(&format!("{}\0", message)[..]);
@@ -270,28 +312,18 @@ where
res.put_i32(len);
res.put_slice(&set_complete[..]);
write_all_half(stream, res).await?;
ready_for_query(stream).await
// ReadyForQuery (idle)
res.put_u8(b'Z');
res.put_i32(5);
res.put_u8(b'I');
write_all_half(stream, res).await
}
/// Send a custom error message to the client.
/// Tell the client we are ready for the next query and no rollback is necessary.
/// Docs on error codes: <https://www.postgresql.org/docs/12/errcodes-appendix.html>.
pub async fn error_response<S>(stream: &mut S, message: &str) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
error_response_terminal(stream, message).await?;
ready_for_query(stream).await
}
/// Send a custom error message to the client.
/// Tell the client we are ready for the next query and no rollback is necessary.
/// Docs on error codes: <https://www.postgresql.org/docs/12/errcodes-appendix.html>.
pub async fn error_response_terminal<S>(stream: &mut S, message: &str) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
/// Docs on error codes: https://www.postgresql.org/docs/12/errcodes-appendix.html
pub async fn error_response(stream: &mut OwnedWriteHalf, message: &str) -> Result<(), Error> {
let mut error = BytesMut::new();
// Error level
@@ -313,57 +345,31 @@ where
// No more fields follow.
error.put_u8(0);
// Ready for query, no rollback needed (I = idle).
let mut ready_for_query = BytesMut::new();
ready_for_query.put_u8(b'Z');
ready_for_query.put_i32(5);
ready_for_query.put_u8(b'I');
// Compose the two message reply.
let mut res = BytesMut::with_capacity(error.len() + 5);
let mut res = BytesMut::with_capacity(error.len() + ready_for_query.len() + 5);
res.put_u8(b'E');
res.put_i32(error.len() as i32 + 4);
res.put(error);
res.put(ready_for_query);
Ok(write_all_half(stream, res).await?)
}
pub async fn wrong_password<S>(stream: &mut S, user: &str) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
let mut error = BytesMut::new();
// Error level
error.put_u8(b'S');
error.put_slice(&b"FATAL\0"[..]);
// Error level (non-translatable)
error.put_u8(b'V');
error.put_slice(&b"FATAL\0"[..]);
// Error code: not sure how much this matters.
error.put_u8(b'C');
error.put_slice(&b"28P01\0"[..]); // system_error, see Appendix A.
// The short error message.
error.put_u8(b'M');
error.put_slice(&format!("password authentication failed for user \"{}\"\0", user).as_bytes());
// No more fields follow.
error.put_u8(0);
// Compose the two message reply.
let mut res = BytesMut::new();
res.put_u8(b'E');
res.put_i32(error.len() as i32 + 4);
res.put(error);
write_all(stream, res).await
}
/// Respond to a SHOW SHARD command.
pub async fn show_response<S>(stream: &mut S, name: &str, value: &str) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
pub async fn show_response(
stream: &mut OwnedWriteHalf,
name: &str,
value: &str,
) -> Result<(), Error> {
// A SELECT response consists of:
// 1. RowDescription
// 2. One or more DataRow
@@ -382,8 +388,12 @@ where
// CommandComplete
res.put(command_complete("SELECT 1"));
write_all_half(stream, res).await?;
ready_for_query(stream).await
// ReadyForQuery
res.put_u8(b'Z');
res.put_i32(5);
res.put_u8(b'I');
write_all_half(stream, res).await
}
pub fn row_description(columns: &Vec<(&str, DataType)>) -> BytesMut {
@@ -429,7 +439,6 @@ pub fn row_description(columns: &Vec<(&str, DataType)>) -> BytesMut {
res
}
/// Create a DataRow message.
pub fn data_row(row: &Vec<String>) -> BytesMut {
let mut res = BytesMut::new();
let mut data_row = BytesMut::new();
@@ -449,7 +458,6 @@ pub fn data_row(row: &Vec<String>) -> BytesMut {
res
}
/// Create a CommandComplete message.
pub fn command_complete(command: &str) -> BytesMut {
let cmd = BytesMut::from(format!("{}\0", command).as_bytes());
let mut res = BytesMut::new();
@@ -460,10 +468,7 @@ pub fn command_complete(command: &str) -> BytesMut {
}
/// Write all data in the buffer to the TcpStream.
pub async fn write_all<S>(stream: &mut S, buf: BytesMut) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
pub async fn write_all(stream: &mut TcpStream, buf: BytesMut) -> Result<(), Error> {
match stream.write_all(&buf).await {
Ok(_) => Ok(()),
Err(_) => return Err(Error::SocketError),
@@ -471,10 +476,7 @@ where
}
/// Write all the data in the buffer to the TcpStream, write owned half (see mpsc).
pub async fn write_all_half<S>(stream: &mut S, buf: BytesMut) -> Result<(), Error>
where
S: tokio::io::AsyncWrite + std::marker::Unpin,
{
pub async fn write_all_half(stream: &mut OwnedWriteHalf, buf: BytesMut) -> Result<(), Error> {
match stream.write_all(&buf).await {
Ok(_) => Ok(()),
Err(_) => return Err(Error::SocketError),
@@ -482,10 +484,7 @@ where
}
/// Read a complete message from the socket.
pub async fn read_message<S>(stream: &mut S) -> Result<BytesMut, Error>
where
S: tokio::io::AsyncRead + std::marker::Unpin,
{
pub async fn read_message(stream: &mut BufReader<OwnedReadHalf>) -> Result<BytesMut, Error> {
let code = match stream.read_u8().await {
Ok(code) => code,
Err(_) => return Err(Error::SocketError),
@@ -511,20 +510,3 @@ where
Ok(bytes)
}
pub fn server_paramater_message(key: &str, value: &str) -> BytesMut {
let mut server_info = BytesMut::new();
let null_byte_size = 1;
let len: usize =
mem::size_of::<i32>() + key.len() + null_byte_size + value.len() + null_byte_size;
server_info.put_slice("S".as_bytes());
server_info.put_i32(len.try_into().unwrap());
server_info.put_slice(key.as_bytes());
server_info.put_bytes(0, 1);
server_info.put_slice(value.as_bytes());
server_info.put_bytes(0, 1);
return server_info;
}

View File

@@ -1,256 +1,116 @@
use arc_swap::ArcSwap;
/// Pooling and failover and banlist.
use async_trait::async_trait;
use bb8::{ManageConnection, Pool, PooledConnection};
use bytes::BytesMut;
use chrono::naive::NaiveDateTime;
use log::{debug, error, info, warn};
use once_cell::sync::Lazy;
use parking_lot::{Mutex, RwLock};
use rand::seq::SliceRandom;
use rand::thread_rng;
use crate::config::{get_config, Address, Role, User};
use crate::errors::Error;
use crate::server::Server;
use crate::stats::Reporter;
use std::collections::HashMap;
use std::sync::Arc;
use std::time::Instant;
use crate::config::{get_config, Address, Role, User};
use crate::errors::Error;
use crate::server::Server;
use crate::sharding::ShardingFunction;
use crate::stats::{get_reporter, Reporter};
// Banlist: bad servers go in here.
pub type BanList = Arc<RwLock<Vec<HashMap<Address, NaiveDateTime>>>>;
pub type ClientServerMap = Arc<Mutex<HashMap<(i32, i32), (i32, i32, String, u16)>>>;
pub type PoolMap = HashMap<(String, String), ConnectionPool>;
/// The connection pool, globally available.
/// This is atomic and safe and read-optimized.
/// The pool is recreated dynamically when the config is reloaded.
pub static POOLS: Lazy<ArcSwap<PoolMap>> = Lazy::new(|| ArcSwap::from_pointee(HashMap::default()));
pub type ClientServerMap = Arc<Mutex<HashMap<(i32, i32), (i32, i32, String, String)>>>;
/// Pool mode:
/// - transaction: server serves one transaction,
/// - session: server is attached to the client.
#[derive(Debug, Clone, Copy, PartialEq)]
pub enum PoolMode {
Session,
Transaction,
}
impl std::fmt::Display for PoolMode {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
match *self {
PoolMode::Session => write!(f, "session"),
PoolMode::Transaction => write!(f, "transaction"),
}
}
}
/// Pool settings.
#[derive(Clone, Debug)]
pub struct PoolSettings {
/// Transaction or Session.
pub pool_mode: PoolMode,
// Number of shards.
pub shards: usize,
// Connecting user.
pub user: User,
// Default server role to connect to.
pub default_role: Option<Role>,
// Enable/disable query parser.
pub query_parser_enabled: bool,
// Read from the primary as well or not.
pub primary_reads_enabled: bool,
// Sharding function.
pub sharding_function: ShardingFunction,
}
impl Default for PoolSettings {
fn default() -> PoolSettings {
PoolSettings {
pool_mode: PoolMode::Transaction,
shards: 1,
user: User::default(),
default_role: None,
query_parser_enabled: false,
primary_reads_enabled: true,
sharding_function: ShardingFunction::PgBigintHash,
}
}
}
/// The globally accessible connection pool.
#[derive(Clone, Debug, Default)]
pub struct ConnectionPool {
/// The pools handled internally by bb8.
databases: Vec<Vec<Pool<ServerPool>>>,
/// The addresses (host, port, role) to handle
/// failover and load balancing deterministically.
addresses: Vec<Vec<Address>>,
/// List of banned addresses (see above)
/// that should not be queried.
round_robin: usize,
banlist: BanList,
/// The statistics aggregator runs in a separate task
/// and receives stats from clients, servers, and the pool.
stats: Reporter,
/// The server information (K messages) have to be passed to the
/// clients on startup. We pre-connect to all shards and replicas
/// on pool creation and save the K messages here.
server_info: BytesMut,
/// Pool configuration.
pub settings: PoolSettings,
}
impl ConnectionPool {
/// Construct the connection pool from the configuration.
pub async fn from_config(client_server_map: ClientServerMap) -> Result<(), Error> {
/// Construct the connection pool from a config file.
pub async fn from_config(
client_server_map: ClientServerMap,
stats: Reporter,
) -> ConnectionPool {
let config = get_config();
let mut shards = Vec::new();
let mut addresses = Vec::new();
let mut banlist = Vec::new();
let mut shard_ids = config
.shards
.clone()
.into_keys()
.map(|x| x.to_string())
.collect::<Vec<String>>();
shard_ids.sort_by_key(|k| k.parse::<i64>().unwrap());
let mut new_pools = HashMap::new();
let mut address_id = 0;
for shard_idx in shard_ids {
let shard = &config.shards[&shard_idx];
let mut pools = Vec::new();
let mut servers = Vec::new();
let mut replica_number = 0;
for (pool_name, pool_config) in &config.pools {
// There is one pool per database/user pair.
for (_, user) in &pool_config.users {
let mut shards = Vec::new();
let mut addresses = Vec::new();
let mut banlist = Vec::new();
let mut shard_ids = pool_config
.shards
.clone()
.into_keys()
.map(|x| x.to_string())
.collect::<Vec<String>>();
// Sort by shard number to ensure consistency.
shard_ids.sort_by_key(|k| k.parse::<i64>().unwrap());
for shard_idx in &shard_ids {
let shard = &pool_config.shards[shard_idx];
let mut pools = Vec::new();
let mut servers = Vec::new();
let mut address_index = 0;
let mut replica_number = 0;
for server in shard.servers.iter() {
let role = match server.2.as_ref() {
"primary" => Role::Primary,
"replica" => Role::Replica,
_ => {
error!("Config error: server role can be 'primary' or 'replica', have: '{}'. Defaulting to 'replica'.", server.2);
Role::Replica
}
};
let address = Address {
id: address_id,
database: shard.database.clone(),
search_path: shard.search_path.clone(),
host: server.0.clone(),
port: server.1 as u16,
role: role,
address_index,
replica_number,
shard: shard_idx.parse::<usize>().unwrap(),
username: user.username.clone(),
pool_name: pool_name.clone(),
};
address_id += 1;
address_index += 1;
if role == Role::Replica {
replica_number += 1;
}
let manager = ServerPool::new(
address.clone(),
user.clone(),
&shard.database,
client_server_map.clone(),
get_reporter(),
);
let pool = Pool::builder()
.max_size(user.pool_size)
.connection_timeout(std::time::Duration::from_millis(
config.general.connect_timeout,
))
.test_on_check_out(false)
.build(manager)
.await
.unwrap();
pools.push(pool);
servers.push(address);
for server in shard.servers.iter() {
let role = match server.2.as_ref() {
"primary" => Role::Primary,
"replica" => Role::Replica,
_ => {
error!("Config error: server role can be 'primary' or 'replica', have: '{}'. Defaulting to 'replica'.", server.2);
Role::Replica
}
};
shards.push(pools);
addresses.push(servers);
banlist.push(HashMap::new());
let address = Address {
host: server.0.clone(),
port: server.1.to_string(),
role: role,
replica_number,
shard: shard_idx.parse::<usize>().unwrap(),
};
if role == Role::Replica {
replica_number += 1;
}
assert_eq!(shards.len(), addresses.len());
let manager = ServerPool::new(
address.clone(),
config.user.clone(),
&shard.database,
client_server_map.clone(),
stats.clone(),
);
let mut pool = ConnectionPool {
databases: shards,
addresses: addresses,
banlist: Arc::new(RwLock::new(banlist)),
stats: get_reporter(),
server_info: BytesMut::new(),
settings: PoolSettings {
pool_mode: match pool_config.pool_mode.as_str() {
"transaction" => PoolMode::Transaction,
"session" => PoolMode::Session,
_ => unreachable!(),
},
// shards: pool_config.shards.clone(),
shards: shard_ids.len(),
user: user.clone(),
default_role: match pool_config.default_role.as_str() {
"any" => None,
"replica" => Some(Role::Replica),
"primary" => Some(Role::Primary),
_ => unreachable!(),
},
query_parser_enabled: pool_config.query_parser_enabled.clone(),
primary_reads_enabled: pool_config.primary_reads_enabled,
sharding_function: match pool_config.sharding_function.as_str() {
"pg_bigint_hash" => ShardingFunction::PgBigintHash,
"sha1" => ShardingFunction::Sha1,
_ => unreachable!(),
},
},
};
let pool = Pool::builder()
.max_size(config.general.pool_size)
.connection_timeout(std::time::Duration::from_millis(
config.general.connect_timeout,
))
.test_on_check_out(false)
.build(manager)
.await
.unwrap();
// Connect to the servers to make sure pool configuration is valid
// before setting it globally.
match pool.validate().await {
Ok(_) => (),
Err(err) => {
error!("Could not validate connection pool: {:?}", err);
return Err(err);
}
};
// There is one pool per database/user pair.
new_pools.insert((pool_name.clone(), user.username.clone()), pool);
pools.push(pool);
servers.push(address);
}
shards.push(pools);
addresses.push(servers);
banlist.push(HashMap::new());
}
POOLS.store(Arc::new(new_pools.clone()));
assert_eq!(shards.len(), addresses.len());
let address_len = addresses.len();
Ok(())
ConnectionPool {
databases: shards,
addresses: addresses,
round_robin: rand::random::<usize>() % address_len, // Start at a random replica
banlist: Arc::new(RwLock::new(banlist)),
stats: stats,
}
}
/// Connect to all shards and grab server information.
@@ -258,11 +118,12 @@ impl ConnectionPool {
/// when they connect.
/// This also warms up the pool for clients that connect when
/// the pooler starts up.
async fn validate(&mut self) -> Result<(), Error> {
pub async fn validate(&mut self) -> Result<BytesMut, Error> {
let mut server_infos = Vec::new();
for shard in 0..self.shards() {
for server in 0..self.servers(shard) {
let connection = match self.databases[shard][server].get().await {
for _ in 0..self.servers(shard) {
let connection = match self.get(shard, None).await {
Ok(conn) => conn,
Err(err) => {
error!("Shard {} down or misconfigured: {:?}", shard, err);
@@ -270,8 +131,10 @@ impl ConnectionPool {
}
};
let proxy = connection;
let server = &*proxy;
let mut proxy = connection.0;
let address = connection.1;
let server = &mut *proxy;
let server_info = server.server_info();
if server_infos.len() > 0 {
@@ -279,7 +142,7 @@ impl ConnectionPool {
if server_info != server_infos[server_infos.len() - 1] {
warn!(
"{:?} has different server configuration than the last server",
proxy.address()
address
);
}
}
@@ -294,160 +157,141 @@ impl ConnectionPool {
return Err(Error::AllServersDown);
}
// We're assuming all servers are identical.
// TODO: not true.
self.server_info = server_infos[0].clone();
Ok(())
Ok(server_infos[0].clone())
}
/// Get a connection from the pool.
pub async fn get(
&self,
shard: usize, // shard number
role: Option<Role>, // primary or replica
process_id: i32, // client id
&mut self,
shard: usize,
role: Option<Role>,
) -> Result<(PooledConnection<'_, ServerPool>, Address), Error> {
let now = Instant::now();
let mut candidates: Vec<&Address> = self.addresses[shard]
.iter()
.filter(|address| address.role == role)
.collect();
let addresses = &self.addresses[shard];
// Random load balancing
candidates.shuffle(&mut thread_rng());
let mut allowed_attempts = match role {
// Primary-specific queries get one attempt, if the primary is down,
// nothing we should do about it I think. It's dangerous to retry
// write queries.
Some(Role::Primary) => 1,
let healthcheck_timeout = get_config().general.healthcheck_timeout;
let healthcheck_delay = get_config().general.healthcheck_delay as u128;
// Replicas get to try as many times as there are replicas
// and connections in the pool.
_ => addresses.len(),
};
while !candidates.is_empty() {
// Get the next candidate
let address = match candidates.pop() {
Some(address) => address,
None => break,
};
debug!("Allowed attempts for {:?}: {}", role, allowed_attempts);
if self.is_banned(&address, role) {
debug!("Address {:?} is banned", address);
let exists = match role {
Some(role) => addresses.iter().filter(|addr| addr.role == role).count() > 0,
None => true,
};
if !exists {
error!("Requested role {:?}, but none are configured", role);
return Err(Error::BadConfig);
}
while allowed_attempts > 0 {
// Round-robin each client's queries.
// If a client only sends one query and then disconnects, it doesn't matter
// which replica it'll go to.
self.round_robin += 1;
let index = self.round_robin % addresses.len();
let address = &addresses[index];
// Make sure you're getting a primary or a replica
// as per request. If no specific role is requested, the first
// available will be chosen.
if address.role != role {
continue;
}
// Indicate we're waiting on a server connection from a pool.
self.stats.client_waiting(process_id, address.id);
allowed_attempts -= 1;
if self.is_banned(address, shard, role) {
continue;
}
// Check if we can connect
let mut conn = match self.databases[address.shard][address.address_index]
.get()
.await
{
let mut conn = match self.databases[shard][index].get().await {
Ok(conn) => conn,
Err(err) => {
error!("Banning instance {:?}, error: {:?}", address, err);
self.ban(&address, process_id);
self.stats
.checkout_time(now.elapsed().as_micros(), process_id, address.id);
error!("Banning replica {}, error: {:?}", index, err);
self.ban(address, shard);
continue;
}
};
// // Check if this server is alive with a health check.
// // Check if this server is alive with a health check
let server = &mut *conn;
let healthcheck_timeout = get_config().general.healthcheck_timeout;
// Will return error if timestamp is greater than current system time, which it should never be set to
let require_healthcheck =
server.last_activity().elapsed().unwrap().as_millis() > healthcheck_delay;
// Do not issue a health check unless it's been a little while
// since we last checked the server is ok.
// Health checks are pretty expensive.
if !require_healthcheck {
self.stats
.checkout_time(now.elapsed().as_micros(), process_id, address.id);
self.stats.server_active(conn.process_id(), address.id);
return Ok((conn, address.clone()));
}
debug!("Running health check on server {:?}", address);
self.stats.server_tested(server.process_id(), address.id);
self.stats.server_tested(server.process_id());
match tokio::time::timeout(
tokio::time::Duration::from_millis(healthcheck_timeout),
server.query(";"), // Cheap query (query parser not used in PG)
server.query("SELECT 1"),
)
.await
{
// Check if health check succeeded.
// Check if health check succeeded
Ok(res) => match res {
Ok(_) => {
self.stats
.checkout_time(now.elapsed().as_micros(), process_id, address.id);
self.stats.server_active(conn.process_id(), address.id);
self.stats.checkout_time(now.elapsed().as_micros());
self.stats.server_idle(conn.process_id());
return Ok((conn, address.clone()));
}
// Health check failed.
Err(err) => {
error!(
"Banning instance {:?} because of failed health check, {:?}",
address, err
);
Err(_) => {
error!("Banning replica {} because of failed health check", index);
// Don't leave a bad connection in the pool.
server.mark_bad();
self.ban(&address, process_id);
self.ban(address, shard);
continue;
}
},
// Health check timed out.
Err(err) => {
error!(
"Banning instance {:?} because of health check timeout, {:?}",
address, err
);
// Health check never came back, database is really really down
Err(_) => {
error!("Banning replica {} because of health check timeout", index);
// Don't leave a bad connection in the pool.
server.mark_bad();
self.ban(&address, process_id);
self.ban(address, shard);
continue;
}
}
}
Err(Error::AllServersDown)
return Err(Error::AllServersDown);
}
/// Ban an address (i.e. replica). It no longer will serve
/// traffic for any new transactions. Existing transactions on that replica
/// will finish successfully or error out to the clients.
pub fn ban(&self, address: &Address, process_id: i32) {
self.stats.client_disconnecting(process_id, address.id);
pub fn ban(&self, address: &Address, shard: usize) {
error!("Banning {:?}", address);
let now = chrono::offset::Utc::now().naive_utc();
let mut guard = self.banlist.write();
guard[address.shard].insert(address.clone(), now);
guard[shard].insert(address.clone(), now);
}
/// Clear the replica to receive traffic again. Takes effect immediately
/// for all new transactions.
pub fn _unban(&self, address: &Address) {
pub fn _unban(&self, address: &Address, shard: usize) {
let mut guard = self.banlist.write();
guard[address.shard].remove(address);
guard[shard].remove(address);
}
/// Check if a replica can serve traffic. If all replicas are banned,
/// we unban all of them. Better to try then not to.
pub fn is_banned(&self, address: &Address, role: Option<Role>) -> bool {
pub fn is_banned(&self, address: &Address, shard: usize, role: Option<Role>) -> bool {
let replicas_available = match role {
Some(Role::Replica) => self.addresses[address.shard]
Some(Role::Replica) => self.addresses[shard]
.iter()
.filter(|addr| addr.role == Role::Replica)
.count(),
None => self.addresses[address.shard].len(),
None => self.addresses[shard].len(),
Some(Role::Primary) => return false, // Primary cannot be banned.
};
@@ -456,17 +300,17 @@ impl ConnectionPool {
let guard = self.banlist.read();
// Everything is banned = nothing is banned.
if guard[address.shard].len() == replicas_available {
if guard[shard].len() == replicas_available {
drop(guard);
let mut guard = self.banlist.write();
guard[address.shard].clear();
guard[shard].clear();
drop(guard);
warn!("Unbanning all replicas.");
return false;
}
// I expect this to miss 99.9999% of the time.
match guard[address.shard].get(address) {
match guard[shard].get(address) {
Some(timestamp) => {
let now = chrono::offset::Utc::now().naive_utc();
let config = get_config();
@@ -476,7 +320,7 @@ impl ConnectionPool {
drop(guard);
warn!("Unbanning {:?}", address);
let mut guard = self.banlist.write();
guard[address.shard].remove(address);
guard[shard].remove(address);
false
} else {
debug!("{:?} is banned", address);
@@ -491,18 +335,14 @@ impl ConnectionPool {
}
}
/// Get the number of configured shards.
pub fn shards(&self) -> usize {
self.databases.len()
}
/// Get the number of servers (primary and replicas)
/// configured for a shard.
pub fn servers(&self, shard: usize) -> usize {
self.addresses[shard].len()
}
/// Get the total number of servers (databases) we are connected to.
pub fn databases(&self) -> usize {
let mut databases = 0;
for shard in 0..self.shards() {
@@ -511,22 +351,15 @@ impl ConnectionPool {
databases
}
/// Get pool state for a particular shard server as reported by bb8.
pub fn pool_state(&self, shard: usize, server: usize) -> bb8::State {
self.databases[shard][server].state()
}
/// Get the address information for a shard server.
pub fn address(&self, shard: usize, server: usize) -> &Address {
&self.addresses[shard][server]
}
pub fn server_info(&self) -> BytesMut {
self.server_info.clone()
}
}
/// Wrapper for the bb8 connection pool.
pub struct ServerPool {
address: Address,
user: User,
@@ -562,16 +395,14 @@ impl ManageConnection for ServerPool {
async fn connect(&self) -> Result<Self::Connection, Self::Error> {
info!(
"Creating a new connection to {:?} using user {:?}",
self.address.name(),
self.user.username
self.address, self.user.name
);
// Put a temporary process_id into the stats
// for server login.
let process_id = rand::random::<i32>();
self.stats.server_login(process_id, self.address.id);
self.stats.server_login(process_id);
// Connect to the PostgreSQL server.
match Server::startup(
&self.address,
&self.user,
@@ -583,12 +414,12 @@ impl ManageConnection for ServerPool {
{
Ok(conn) => {
// Remove the temporary process_id from the stats.
self.stats.server_disconnecting(process_id, self.address.id);
self.stats.server_disconnecting(process_id);
Ok(conn)
}
Err(err) => {
// Remove the temporary process_id from the stats.
self.stats.server_disconnecting(process_id, self.address.id);
self.stats.server_disconnecting(process_id);
Err(err)
}
}
@@ -604,24 +435,3 @@ impl ManageConnection for ServerPool {
conn.is_bad()
}
}
/// Get the connection pool
pub fn get_pool(db: String, user: String) -> Option<ConnectionPool> {
match get_all_pools().get(&(db, user)) {
Some(pool) => Some(pool.clone()),
None => None,
}
}
/// How many total servers we have in the config.
pub fn get_number_of_addresses() -> usize {
get_all_pools()
.iter()
.map(|(_, pool)| pool.databases())
.sum()
}
/// Get a pointer to all configured pools.
pub fn get_all_pools() -> HashMap<(String, String), ConnectionPool> {
return (*(*POOLS.load())).clone();
}

View File

@@ -1,210 +0,0 @@
use hyper::service::{make_service_fn, service_fn};
use hyper::{Body, Method, Request, Response, Server, StatusCode};
use log::{error, info, warn};
use phf::phf_map;
use std::collections::HashMap;
use std::fmt;
use std::net::SocketAddr;
use crate::config::Address;
use crate::pool::get_all_pools;
use crate::stats::get_stats;
struct MetricHelpType {
help: &'static str,
ty: &'static str,
}
// reference for metric types: https://prometheus.io/docs/concepts/metric_types/
// counters only increase
// gauges can arbitrarily increase or decrease
static METRIC_HELP_AND_TYPES_LOOKUP: phf::Map<&'static str, MetricHelpType> = phf_map! {
"total_query_count" => MetricHelpType {
help: "Number of queries sent by all clients",
ty: "counter",
},
"total_query_time" => MetricHelpType {
help: "Total amount of time for queries to execute",
ty: "counter",
},
"total_received" => MetricHelpType {
help: "Number of bytes received from the server",
ty: "counter",
},
"total_sent" => MetricHelpType {
help: "Number of bytes sent to the server",
ty: "counter",
},
"total_xact_count" => MetricHelpType {
help: "Total number of transactions started by the client",
ty: "counter",
},
"total_xact_time" => MetricHelpType {
help: "Total amount of time for all transactions to execute",
ty: "counter",
},
"total_wait_time" => MetricHelpType {
help: "Total time client waited for a server connection",
ty: "counter",
},
"avg_query_count" => MetricHelpType {
help: "Average of total_query_count every 15 seconds",
ty: "gauge",
},
"avg_query_time" => MetricHelpType {
help: "Average time taken for queries to execute every 15 seconds",
ty: "gauge",
},
"avg_recv" => MetricHelpType {
help: "Average of total_received bytes every 15 seconds",
ty: "gauge",
},
"avg_sent" => MetricHelpType {
help: "Average of total_sent bytes every 15 seconds",
ty: "gauge",
},
"avg_xact_count" => MetricHelpType {
help: "Average of total_xact_count every 15 seconds",
ty: "gauge",
},
"avg_xact_time" => MetricHelpType {
help: "Average of total_xact_time every 15 seconds",
ty: "gauge",
},
"avg_wait_time" => MetricHelpType {
help: "Average of total_wait_time every 15 seconds",
ty: "gauge",
},
"maxwait_us" => MetricHelpType {
help: "The time a client waited for a server connection in microseconds",
ty: "gauge",
},
"maxwait" => MetricHelpType {
help: "The time a client waited for a server connection in seconds",
ty: "gauge",
},
"cl_waiting" => MetricHelpType {
help: "How many clients are waiting for a connection from the pool",
ty: "gauge",
},
"cl_active" => MetricHelpType {
help: "How many clients are actively communicating with a server",
ty: "gauge",
},
"cl_idle" => MetricHelpType {
help: "How many clients are idle",
ty: "gauge",
},
"sv_idle" => MetricHelpType {
help: "How many server connections are idle",
ty: "gauge",
},
"sv_active" => MetricHelpType {
help: "How many server connections are actively communicating with a client",
ty: "gauge",
},
"sv_login" => MetricHelpType {
help: "How many server connections are currently being created",
ty: "gauge",
},
"sv_tested" => MetricHelpType {
help: "How many server connections are currently waiting on a health check to succeed",
ty: "gauge",
},
};
struct PrometheusMetric {
name: String,
help: String,
ty: String,
labels: HashMap<&'static str, String>,
value: i64,
}
impl fmt::Display for PrometheusMetric {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
let formatted_labels = self
.labels
.iter()
.map(|(key, value)| format!("{}=\"{}\"", key, value))
.collect::<Vec<_>>()
.join(",");
write!(
f,
"# HELP {name} {help}\n# TYPE {name} {ty}\n{name}{{{formatted_labels}}} {value}\n",
name = format_args!("pgcat_{}", self.name),
help = self.help,
ty = self.ty,
formatted_labels = formatted_labels,
value = self.value
)
}
}
impl PrometheusMetric {
fn new(address: &Address, name: &str, value: i64) -> Option<PrometheusMetric> {
let mut labels = HashMap::new();
labels.insert("host", address.host.clone());
labels.insert("shard", address.shard.to_string());
labels.insert("role", address.role.to_string());
labels.insert("database", address.database.to_string());
METRIC_HELP_AND_TYPES_LOOKUP
.get(name)
.map(|metric| PrometheusMetric {
name: name.to_owned(),
help: metric.help.to_owned(),
ty: metric.ty.to_owned(),
labels,
value,
})
}
}
async fn prometheus_stats(request: Request<Body>) -> Result<Response<Body>, hyper::http::Error> {
match (request.method(), request.uri().path()) {
(&Method::GET, "/metrics") => {
let stats = get_stats();
let mut lines = Vec::new();
for (_, pool) in get_all_pools() {
for shard in 0..pool.shards() {
for server in 0..pool.servers(shard) {
let address = pool.address(shard, server);
if let Some(address_stats) = stats.get(&address.id) {
for (key, value) in address_stats.iter() {
if let Some(prometheus_metric) =
PrometheusMetric::new(address, key, *value)
{
lines.push(prometheus_metric.to_string());
} else {
warn!("Metric {} not implemented for {}", key, address.name());
}
}
}
}
}
}
Response::builder()
.header("content-type", "text/plain; version=0.0.4")
.body(lines.join("\n").into())
}
_ => Response::builder()
.status(StatusCode::NOT_FOUND)
.body("".into()),
}
}
pub async fn start_metric_server(http_addr: SocketAddr) {
let http_service_factory =
make_service_fn(|_conn| async { Ok::<_, hyper::Error>(service_fn(prometheus_stats)) });
let server = Server::bind(&http_addr.into()).serve(http_service_factory);
info!(
"Exposing prometheus metrics on http://{}/metrics.",
http_addr
);
if let Err(e) = server.await {
error!("Failed to run HTTP server: {}.", e);
}
}

View File

@@ -1,3 +1,5 @@
use crate::config::{get_config, Role};
use crate::sharding::{Sharder, ShardingFunction};
/// Route queries automatically based on explicitely requested
/// or implied query characteristics.
use bytes::{Buf, BytesMut};
@@ -8,22 +10,14 @@ use sqlparser::ast::Statement::{Query, StartTransaction};
use sqlparser::dialect::PostgreSqlDialect;
use sqlparser::parser::Parser;
use crate::config::Role;
use crate::pool::PoolSettings;
use crate::sharding::Sharder;
/// Regexes used to parse custom commands.
const CUSTOM_SQL_REGEXES: [&str; 7] = [
const CUSTOM_SQL_REGEXES: [&str; 5] = [
r"(?i)^ *SET SHARDING KEY TO '?([0-9]+)'? *;? *$",
r"(?i)^ *SET SHARD TO '?([0-9]+|ANY)'? *;? *$",
r"(?i)^ *SET SHARD TO '?([0-9]+)'? *;? *$",
r"(?i)^ *SHOW SHARD *;? *$",
r"(?i)^ *SET SERVER ROLE TO '(PRIMARY|REPLICA|ANY|AUTO|DEFAULT)' *;? *$",
r"(?i)^ *SHOW SERVER ROLE *;? *$",
r"(?i)^ *SET PRIMARY READS TO '?(on|off|default)'? *;? *$",
r"(?i)^ *SHOW PRIMARY READS *;? *$",
];
/// Custom commands.
#[derive(PartialEq, Debug)]
pub enum Command {
SetShardingKey,
@@ -31,37 +25,39 @@ pub enum Command {
ShowShard,
SetServerRole,
ShowServerRole,
SetPrimaryReads,
ShowPrimaryReads,
}
/// Quickly test for match when a query is received.
// Quick test
static CUSTOM_SQL_REGEX_SET: OnceCell<RegexSet> = OnceCell::new();
// Get the value inside the custom command.
// Capture value
static CUSTOM_SQL_REGEX_LIST: OnceCell<Vec<Regex>> = OnceCell::new();
/// The query router.
pub struct QueryRouter {
/// Which shard we should be talking to right now.
// By default, queries go here, unless we have better information
// about what the client wants.
default_server_role: Option<Role>,
// Number of shards in the cluster.
shards: usize,
// Which shard we should be talking to right now.
active_shard: Option<usize>,
/// Which server should we be talking to.
// Should we be talking to a primary or a replica?
active_role: Option<Role>,
/// Should we try to parse queries to route them to replicas or primary automatically
query_parser_enabled: bool,
/// Include the primary into the replica pool for reads.
// Include the primary into the replica pool?
primary_reads_enabled: bool,
/// Pool configuration.
pool_settings: PoolSettings,
// Should we try to parse queries?
query_parser_enabled: bool,
// Which sharding function are we using?
sharding_function: ShardingFunction,
}
impl QueryRouter {
/// One-time initialization of regexes
/// that parse our custom SQL protocol.
pub fn setup() -> bool {
let set = match RegexSet::new(&CUSTOM_SQL_REGEXES) {
Ok(rgx) => rgx,
@@ -76,7 +72,10 @@ impl QueryRouter {
.map(|rgx| Regex::new(rgx).unwrap())
.collect();
assert_eq!(list.len(), set.len());
// Impossible
if list.len() != set.len() {
return false;
}
match CUSTOM_SQL_REGEX_LIST.set(list) {
Ok(_) => true,
@@ -89,28 +88,38 @@ impl QueryRouter {
}
}
/// Create a new instance of the query router.
/// Each client gets its own.
pub fn new() -> QueryRouter {
QueryRouter {
active_shard: None,
active_role: None,
query_parser_enabled: false,
primary_reads_enabled: false,
pool_settings: PoolSettings::default(),
}
}
let config = get_config();
/// Pool settings can change because of a config reload.
pub fn update_pool_settings(&mut self, pool_settings: PoolSettings) {
self.pool_settings = pool_settings;
let default_server_role = match config.query_router.default_role.as_ref() {
"any" => None,
"primary" => Some(Role::Primary),
"replica" => Some(Role::Replica),
_ => unreachable!(),
};
let sharding_function = match config.query_router.sharding_function.as_ref() {
"pg_bigint_hash" => ShardingFunction::PgBigintHash,
"sha1" => ShardingFunction::Sha1,
_ => unreachable!(),
};
QueryRouter {
default_server_role: default_server_role,
shards: config.shards.len(),
active_role: default_server_role,
active_shard: None,
primary_reads_enabled: config.query_router.primary_reads_enabled,
query_parser_enabled: config.query_router.query_parser_enabled,
sharding_function,
}
}
/// Try to parse a command and execute it.
pub fn try_execute_command(&mut self, mut buf: BytesMut) -> Option<(Command, String)> {
let code = buf.get_u8() as char;
// Only simple protocol supported for commands.
if code != 'Q' {
return None;
}
@@ -130,10 +139,7 @@ impl QueryRouter {
let matches: Vec<_> = regex_set.matches(&query).into_iter().collect();
// This is not a custom query, try to infer which
// server it'll go to if the query parser is enabled.
if matches.len() != 1 {
debug!("Regular query, not a command");
return None;
}
@@ -143,21 +149,17 @@ impl QueryRouter {
2 => Command::ShowShard,
3 => Command::SetServerRole,
4 => Command::ShowServerRole,
5 => Command::SetPrimaryReads,
6 => Command::ShowPrimaryReads,
_ => unreachable!(),
};
let mut value = match command {
Command::SetShardingKey
| Command::SetShard
| Command::SetServerRole
| Command::SetPrimaryReads => {
Command::SetShardingKey | Command::SetShard | Command::SetServerRole => {
// Capture value. I know this re-runs the regex engine, but I haven't
// figured out a better way just yet. I think I can write a single Regex
// that matches all 5 custom SQL patterns, but maybe that's not very legible?
//
// I think this is faster than running the Regex engine 5 times.
// I think this is faster than running the Regex engine 5 times, so
// this is a strong maybe for me so far.
match regex_list[matches[0]].captures(&query) {
Some(captures) => match captures.get(1) {
Some(value) => value.as_str().to_string(),
@@ -179,29 +181,18 @@ impl QueryRouter {
}
}
},
Command::ShowPrimaryReads => match self.primary_reads_enabled {
true => String::from("on"),
false => String::from("off"),
},
};
match command {
Command::SetShardingKey => {
let sharder = Sharder::new(
self.pool_settings.shards,
self.pool_settings.sharding_function,
);
let sharder = Sharder::new(self.shards, self.sharding_function);
let shard = sharder.shard(value.parse::<i64>().unwrap());
self.active_shard = Some(shard);
value = shard.to_string();
}
Command::SetShard => {
self.active_shard = match value.to_ascii_uppercase().as_ref() {
"ANY" => Some(rand::random::<usize>() % self.pool_settings.shards),
_ => Some(value.parse::<usize>().unwrap()),
};
self.active_shard = Some(value.parse::<usize>().unwrap());
}
Command::SetServerRole => {
@@ -227,8 +218,9 @@ impl QueryRouter {
}
"default" => {
self.active_role = self.pool_settings.default_role;
self.query_parser_enabled = self.query_parser_enabled;
// TODO: reset query parser to default here.
self.active_role = self.default_server_role;
self.query_parser_enabled = get_config().query_router.query_parser_enabled;
self.active_role
}
@@ -236,19 +228,6 @@ impl QueryRouter {
};
}
Command::SetPrimaryReads => {
if value == "on" {
debug!("Setting primary reads to on");
self.primary_reads_enabled = true;
} else if value == "off" {
debug!("Setting primary reads to off");
self.primary_reads_enabled = false;
} else if value == "default" {
debug!("Setting primary reads to default");
self.primary_reads_enabled = self.pool_settings.primary_reads_enabled;
}
}
_ => (),
}
@@ -257,20 +236,16 @@ impl QueryRouter {
/// Try to infer which server to connect to based on the contents of the query.
pub fn infer_role(&mut self, mut buf: BytesMut) -> bool {
debug!("Inferring role");
let code = buf.get_u8() as char;
let len = buf.get_i32() as usize;
let query = match code {
// Query
'Q' => {
let query = String::from_utf8_lossy(&buf[..len - 5]).to_string();
debug!("Query: '{}'", query);
query
}
// Parse (prepared statement)
'P' => {
let mut start = 0;
let mut end;
@@ -293,7 +268,6 @@ impl QueryRouter {
query.replace("$", "") // Remove placeholders turning them into "values"
}
_ => return false,
};
@@ -345,23 +319,28 @@ impl QueryRouter {
}
}
pub fn set_shard(&mut self, shard: usize) {
self.active_shard = Some(shard);
/// Reset the router back to defaults.
/// This must be called at the end of every transaction in transaction mode.
pub fn _reset(&mut self) {
self.active_role = self.default_server_role;
self.active_shard = None;
}
/// Should we attempt to parse queries?
#[allow(dead_code)]
pub fn query_parser_enabled(&self) -> bool {
self.query_parser_enabled
}
#[allow(dead_code)]
pub fn toggle_primary_reads(&mut self, value: bool) {
self.primary_reads_enabled = value;
}
}
#[cfg(test)]
mod test {
use super::*;
use crate::messages::simple_query;
use crate::pool::PoolMode;
use crate::sharding::ShardingFunction;
use bytes::BufMut;
#[test]
@@ -378,8 +357,7 @@ mod test {
let mut qr = QueryRouter::new();
assert!(qr.try_execute_command(simple_query("SET SERVER ROLE TO 'auto'")) != None);
assert_eq!(qr.query_parser_enabled(), true);
assert!(qr.try_execute_command(simple_query("SET PRIMARY READS TO off")) != None);
qr.toggle_primary_reads(false);
let queries = vec![
simple_query("SELECT * FROM items WHERE id = 5"),
@@ -420,7 +398,7 @@ mod test {
QueryRouter::setup();
let mut qr = QueryRouter::new();
let query = simple_query("SELECT * FROM items WHERE id = 5");
assert!(qr.try_execute_command(simple_query("SET PRIMARY READS TO on")) != None);
qr.toggle_primary_reads(true);
assert!(qr.infer_role(query));
assert_eq!(qr.role(), None);
@@ -431,7 +409,7 @@ mod test {
QueryRouter::setup();
let mut qr = QueryRouter::new();
qr.try_execute_command(simple_query("SET SERVER ROLE TO 'auto'"));
assert!(qr.try_execute_command(simple_query("SET PRIMARY READS TO off")) != None);
qr.toggle_primary_reads(false);
let prepared_stmt = BytesMut::from(
&b"WITH t AS (SELECT * FROM items WHERE name = $1) SELECT * FROM t WHERE id = $2\0"[..],
@@ -460,10 +438,6 @@ mod test {
"SET SERVER ROLE TO 'any'",
"SET SERVER ROLE TO 'auto'",
"SHOW SERVER ROLE",
"SET PRIMARY READS TO 'on'",
"SET PRIMARY READS TO 'off'",
"SET PRIMARY READS TO 'default'",
"SHOW PRIMARY READS",
// Lower case
"set sharding key to '1'",
"set shard to '1'",
@@ -473,13 +447,9 @@ mod test {
"set server role to 'any'",
"set server role to 'auto'",
"show server role",
"set primary reads to 'on'",
"set primary reads to 'OFF'",
"set primary reads to 'deFaUlt'",
// No quotes
"SET SHARDING KEY TO 11235",
"SET SHARD TO 15",
"SET PRIMARY READS TO off",
// Spaces and semicolon
" SET SHARDING KEY TO 11235 ; ",
" SET SHARD TO 15; ",
@@ -487,23 +457,18 @@ mod test {
" SET SERVER ROLE TO 'primary'; ",
" SET SERVER ROLE TO 'primary' ; ",
" SET SERVER ROLE TO 'primary' ;",
" SET PRIMARY READS TO 'off' ;",
];
// Which regexes it'll match to in the list
let matches = [
0, 1, 2, 3, 3, 3, 3, 4, 5, 5, 5, 6, 0, 1, 2, 3, 3, 3, 3, 4, 5, 5, 5, 0, 1, 5, 0, 1, 0,
3, 3, 3, 5,
0, 1, 2, 3, 3, 3, 3, 4, 0, 1, 2, 3, 3, 3, 3, 4, 0, 1, 0, 1, 0, 3, 3, 3,
];
let list = CUSTOM_SQL_REGEX_LIST.get().unwrap();
let set = CUSTOM_SQL_REGEX_SET.get().unwrap();
for (i, test) in tests.iter().enumerate() {
if !list[matches[i]].is_match(test) {
println!("{} does not match {}", test, list[matches[i]]);
assert!(false);
}
assert!(list[matches[i]].is_match(test));
assert_eq!(set.matches(test).into_iter().collect::<Vec<_>>().len(), 1);
}
@@ -526,9 +491,9 @@ mod test {
let query = simple_query("SET SHARDING KEY TO 13");
assert_eq!(
qr.try_execute_command(query),
Some((Command::SetShardingKey, String::from("0")))
Some((Command::SetShardingKey, String::from("1")))
);
assert_eq!(qr.shard(), 0);
assert_eq!(qr.shard(), 1);
// SetShard
let query = simple_query("SET SHARD TO '1'");
@@ -572,26 +537,6 @@ mod test {
Some((Command::ShowServerRole, String::from(*role)))
);
}
let primary_reads = ["on", "off", "default"];
let primary_reads_enabled = ["on", "off", "on"];
for (idx, primary_reads) in primary_reads.iter().enumerate() {
assert_eq!(
qr.try_execute_command(simple_query(&format!(
"SET PRIMARY READS TO {}",
primary_reads
))),
Some((Command::SetPrimaryReads, String::from(*primary_reads)))
);
assert_eq!(
qr.try_execute_command(simple_query("SHOW PRIMARY READS")),
Some((
Command::ShowPrimaryReads,
String::from(primary_reads_enabled[idx])
))
);
}
}
#[test]
@@ -599,7 +544,7 @@ mod test {
QueryRouter::setup();
let mut qr = QueryRouter::new();
let query = simple_query("SET SERVER ROLE TO 'auto'");
assert!(qr.try_execute_command(simple_query("SET PRIMARY READS TO off")) != None);
qr.toggle_primary_reads(false);
assert!(qr.try_execute_command(query) != None);
assert!(qr.query_parser_enabled());
@@ -616,42 +561,6 @@ mod test {
assert!(qr.query_parser_enabled());
let query = simple_query("SET SERVER ROLE TO 'default'");
assert!(qr.try_execute_command(query) != None);
assert!(qr.query_parser_enabled());
}
#[test]
fn test_update_from_pool_settings() {
QueryRouter::setup();
let pool_settings = PoolSettings {
pool_mode: PoolMode::Transaction,
shards: 0,
user: crate::config::User::default(),
default_role: Some(Role::Replica),
query_parser_enabled: true,
primary_reads_enabled: false,
sharding_function: ShardingFunction::PgBigintHash,
};
let mut qr = QueryRouter::new();
assert_eq!(qr.active_role, None);
assert_eq!(qr.active_shard, None);
assert_eq!(qr.query_parser_enabled, false);
assert_eq!(qr.primary_reads_enabled, false);
// Internal state must not be changed due to this, only defaults
qr.update_pool_settings(pool_settings.clone());
assert_eq!(qr.active_role, None);
assert_eq!(qr.active_shard, None);
assert_eq!(qr.query_parser_enabled, false);
assert_eq!(qr.primary_reads_enabled, false);
let q1 = simple_query("SET SERVER ROLE TO 'primary'");
assert!(qr.try_execute_command(q1) != None);
assert_eq!(qr.active_role.unwrap(), Role::Primary);
let q2 = simple_query("SET SERVER ROLE TO 'default'");
assert!(qr.try_execute_command(q2) != None);
assert_eq!(qr.active_role.unwrap(), pool_settings.clone().default_role);
assert!(!qr.query_parser_enabled());
}
}

View File

@@ -1,320 +0,0 @@
// SCRAM-SHA-256 authentication. Heavily inspired by
// https://github.com/sfackler/rust-postgres/
// SASL implementation.
use bytes::BytesMut;
use hmac::{Hmac, Mac};
use rand::{self, Rng};
use sha2::digest::FixedOutput;
use sha2::{Digest, Sha256};
use std::fmt::Write;
use crate::constants::*;
use crate::errors::Error;
/// Normalize a password string. Postgres
/// passwords don't have to be UTF-8.
fn normalize(pass: &[u8]) -> Vec<u8> {
let pass = match std::str::from_utf8(pass) {
Ok(pass) => pass,
Err(_) => return pass.to_vec(),
};
match stringprep::saslprep(pass) {
Ok(pass) => pass.into_owned().into_bytes(),
Err(_) => pass.as_bytes().to_vec(),
}
}
/// Keep the SASL state through the exchange.
/// It takes 3 messages to complete the authentication.
pub struct ScramSha256 {
password: String,
salted_password: [u8; 32],
auth_message: String,
message: BytesMut,
nonce: String,
}
impl ScramSha256 {
/// Create the Scram state from a password. It'll automatically
/// generate a nonce.
pub fn new(password: &str) -> ScramSha256 {
let mut rng = rand::thread_rng();
let nonce = (0..NONCE_LENGTH)
.map(|_| {
let mut v = rng.gen_range(0x21u8..0x7e);
if v == 0x2c {
v = 0x7e
}
v as char
})
.collect::<String>();
Self::from_nonce(password, &nonce)
}
/// Used for testing.
pub fn from_nonce(password: &str, nonce: &str) -> ScramSha256 {
let message = BytesMut::from(&format!("{}n=,r={}", "n,,", nonce).as_bytes()[..]);
ScramSha256 {
password: password.to_string(),
nonce: String::from(nonce),
message,
salted_password: [0u8; 32],
auth_message: String::new(),
}
}
/// Get the current state of the SASL authentication.
pub fn message(&mut self) -> BytesMut {
self.message.clone()
}
/// Update the state with message received from server.
pub fn update(&mut self, message: &BytesMut) -> Result<BytesMut, Error> {
let server_message = Message::parse(message)?;
if !server_message.nonce.starts_with(&self.nonce) {
return Err(Error::ProtocolSyncError);
}
let salt = match base64::decode(&server_message.salt) {
Ok(salt) => salt,
Err(_) => return Err(Error::ProtocolSyncError),
};
let salted_password = Self::hi(
&normalize(&self.password.as_bytes()[..]),
&salt,
server_message.iterations,
);
// Save for verification of final server message.
self.salted_password = salted_password;
let mut hmac = match Hmac::<Sha256>::new_from_slice(&salted_password) {
Ok(hmac) => hmac,
Err(_) => return Err(Error::ServerError),
};
hmac.update(b"Client Key");
let client_key = hmac.finalize().into_bytes();
let mut hash = Sha256::default();
hash.update(client_key.as_slice());
let stored_key = hash.finalize_fixed();
let mut cbind_input = vec![];
cbind_input.extend("n,,".as_bytes());
let cbind_input = base64::encode(&cbind_input);
self.message.clear();
// Start writing the client reply.
match write!(
&mut self.message,
"c={},r={}",
cbind_input, server_message.nonce
) {
Ok(_) => (),
Err(_) => return Err(Error::ServerError),
};
let auth_message = format!(
"n=,r={},{},{}",
self.nonce,
String::from_utf8_lossy(&message[..]),
String::from_utf8_lossy(&self.message[..])
);
let mut hmac = match Hmac::<Sha256>::new_from_slice(&stored_key) {
Ok(hmac) => hmac,
Err(_) => return Err(Error::ServerError),
};
hmac.update(auth_message.as_bytes());
// Save the auth message for server final message verification.
self.auth_message = auth_message;
let client_signature = hmac.finalize().into_bytes();
// Sign the client proof.
let mut client_proof = client_key;
for (proof, signature) in client_proof.iter_mut().zip(client_signature) {
*proof ^= signature;
}
match write!(&mut self.message, ",p={}", base64::encode(&*client_proof)) {
Ok(_) => (),
Err(_) => return Err(Error::ServerError),
};
Ok(self.message.clone())
}
/// Verify final server message.
pub fn finish(&mut self, message: &BytesMut) -> Result<(), Error> {
let final_message = FinalMessage::parse(message)?;
let verifier = match base64::decode(&final_message.value) {
Ok(verifier) => verifier,
Err(_) => return Err(Error::ProtocolSyncError),
};
let mut hmac = match Hmac::<Sha256>::new_from_slice(&self.salted_password) {
Ok(hmac) => hmac,
Err(_) => return Err(Error::ServerError),
};
hmac.update(b"Server Key");
let server_key = hmac.finalize().into_bytes();
let mut hmac = match Hmac::<Sha256>::new_from_slice(&server_key) {
Ok(hmac) => hmac,
Err(_) => return Err(Error::ServerError),
};
hmac.update(self.auth_message.as_bytes());
match hmac.verify_slice(&verifier) {
Ok(_) => Ok(()),
Err(_) => return Err(Error::ServerError),
}
}
/// Hash the password with the salt i-times.
fn hi(str: &[u8], salt: &[u8], i: u32) -> [u8; 32] {
let mut hmac =
Hmac::<Sha256>::new_from_slice(str).expect("HMAC is able to accept all key sizes");
hmac.update(salt);
hmac.update(&[0, 0, 0, 1]);
let mut prev = hmac.finalize().into_bytes();
let mut hi = prev;
for _ in 1..i {
let mut hmac = Hmac::<Sha256>::new_from_slice(str).expect("already checked above");
hmac.update(&prev);
prev = hmac.finalize().into_bytes();
for (hi, prev) in hi.iter_mut().zip(prev) {
*hi ^= prev;
}
}
hi.into()
}
}
/// Parse the server challenge.
struct Message {
nonce: String,
salt: String,
iterations: u32,
}
impl Message {
/// Parse the server SASL challenge.
fn parse(message: &BytesMut) -> Result<Message, Error> {
let parts = String::from_utf8_lossy(&message[..])
.split(",")
.map(|s| s.to_string())
.collect::<Vec<String>>();
if parts.len() != 3 {
return Err(Error::ProtocolSyncError);
}
let nonce = str::replace(&parts[0], "r=", "");
let salt = str::replace(&parts[1], "s=", "");
let iterations = match str::replace(&parts[2], "i=", "").parse::<u32>() {
Ok(iterations) => iterations,
Err(_) => return Err(Error::ProtocolSyncError),
};
Ok(Message {
nonce,
salt,
iterations,
})
}
}
/// Parse server final validation message.
struct FinalMessage {
value: String,
}
impl FinalMessage {
/// Parse the server final validation message.
pub fn parse(message: &BytesMut) -> Result<FinalMessage, Error> {
if !message.starts_with(b"v=") || message.len() < 4 {
return Err(Error::ProtocolSyncError);
}
Ok(FinalMessage {
value: String::from_utf8_lossy(&message[2..]).to_string(),
})
}
}
#[cfg(test)]
mod test {
use super::*;
#[test]
fn parse_server_first_message() {
let message = BytesMut::from(
&"r=fyko+d2lbbFgONRv9qkxdawL3rfcNHYJY1ZVvWVs7j,s=QSXCR+Q6sek8bf92,i=4096".as_bytes()[..],
);
let message = Message::parse(&message).unwrap();
assert_eq!(message.nonce, "fyko+d2lbbFgONRv9qkxdawL3rfcNHYJY1ZVvWVs7j");
assert_eq!(message.salt, "QSXCR+Q6sek8bf92");
assert_eq!(message.iterations, 4096);
}
#[test]
fn parse_server_last_message() {
let f = FinalMessage::parse(&BytesMut::from(
&"v=U+ppxD5XUKtradnv8e2MkeupiA8FU87Sg8CXzXHDAzw".as_bytes()[..],
))
.unwrap();
assert_eq!(
f.value,
"U+ppxD5XUKtradnv8e2MkeupiA8FU87Sg8CXzXHDAzw".to_string()
);
}
// recorded auth exchange from psql
#[test]
fn exchange() {
let password = "foobar";
let nonce = "9IZ2O01zb9IgiIZ1WJ/zgpJB";
let client_first = "n,,n=,r=9IZ2O01zb9IgiIZ1WJ/zgpJB";
let server_first =
"r=9IZ2O01zb9IgiIZ1WJ/zgpJBjx/oIRLs02gGSHcw1KEty3eY,s=fs3IXBy7U7+IvVjZ,i\
=4096";
let client_final =
"c=biws,r=9IZ2O01zb9IgiIZ1WJ/zgpJBjx/oIRLs02gGSHcw1KEty3eY,p=AmNKosjJzS3\
1NTlQYNs5BTeQjdHdk7lOflDo5re2an8=";
let server_final = "v=U+ppxD5XUKtradnv8e2MkeupiA8FU87Sg8CXzXHDAzw=";
let mut scram = ScramSha256::from_nonce(password, nonce);
let message = scram.message();
assert_eq!(std::str::from_utf8(&message).unwrap(), client_first);
let result = scram
.update(&BytesMut::from(&server_first.as_bytes()[..]))
.unwrap();
assert_eq!(std::str::from_utf8(&result).unwrap(), client_final);
scram
.finish(&BytesMut::from(&server_final.as_bytes()[..]))
.unwrap();
}
}

View File

@@ -1,8 +1,7 @@
/// Implementation of the PostgreSQL server (database) protocol.
/// Here we are pretending to the a Postgres client.
use bytes::{Buf, BufMut, BytesMut};
///! Implementation of the PostgreSQL server (database) protocol.
///! Here we are pretending to the a Postgres client.
use log::{debug, error, info, trace};
use std::time::SystemTime;
use tokio::io::{AsyncReadExt, BufReader};
use tokio::net::{
tcp::{OwnedReadHalf, OwnedWriteHalf},
@@ -13,55 +12,48 @@ use crate::config::{Address, User};
use crate::constants::*;
use crate::errors::Error;
use crate::messages::*;
use crate::scram::ScramSha256;
use crate::stats::Reporter;
use crate::ClientServerMap;
/// Server state.
pub struct Server {
/// Server host, e.g. localhost,
/// port, e.g. 5432, and role, e.g. primary or replica.
// Server host, e.g. localhost,
// port, e.g. 5432, and role, e.g. primary or replica.
address: Address,
/// Buffered read socket.
// Buffered read socket.
read: BufReader<OwnedReadHalf>,
/// Unbuffered write socket (our client code buffers).
// Unbuffered write socket (our client code buffers).
write: OwnedWriteHalf,
/// Our server response buffer. We buffer data before we give it to the client.
// Our server response buffer. We buffer data before we give it to the client.
buffer: BytesMut,
/// Server information the server sent us over on startup.
// Server information the server sent us over on startup.
server_info: BytesMut,
/// Backend id and secret key used for query cancellation.
// Backend id and secret key used for query cancellation.
process_id: i32,
secret_key: i32,
/// Is the server inside a transaction or idle.
// Is the server inside a transaction or idle.
in_transaction: bool,
/// Is there more data for the client to read.
// Is there more data for the client to read.
data_available: bool,
/// Is the server broken? We'll remote it from the pool if so.
// Is the server broken? We'll remote it from the pool if so.
bad: bool,
/// Mapping of clients and servers used for query cancellation.
// Mapping of clients and servers used for query cancellation.
client_server_map: ClientServerMap,
/// Server connected at.
// Server connected at.
connected_at: chrono::naive::NaiveDateTime,
/// Reports various metrics, e.g. data sent & received.
// Reports various metrics, e.g. data sent & received.
stats: Reporter,
/// Application name using the server at the moment.
application_name: String,
// Last time that a successful server send or response happened
last_activity: SystemTime,
}
impl Server {
@@ -75,7 +67,7 @@ impl Server {
stats: Reporter,
) -> Result<Server, Error> {
let mut stream =
match TcpStream::connect(&format!("{}:{}", &address.host, address.port)).await {
match TcpStream::connect(&format!("{}:{}", &address.host, &address.port)).await {
Ok(stream) => stream,
Err(err) => {
error!("Could not connect to server: {}", err);
@@ -85,14 +77,8 @@ impl Server {
trace!("Sending StartupMessage");
// StartupMessage
startup(
&mut stream,
&user.username,
database,
address.search_path.as_ref(),
)
.await?;
// Send the startup packet telling the server we're a normal Postgres client.
startup(&mut stream, &user.name, database).await?;
let mut server_info = BytesMut::new();
let mut process_id: i32 = 0;
@@ -100,8 +86,6 @@ impl Server {
// We'll be handling multiple packets, but they will all be structured the same.
// We'll loop here until this exchange is complete.
let mut scram = ScramSha256::new(&user.password);
loop {
let code = match stream.read_u8().await {
Ok(code) => code as char,
@@ -137,97 +121,12 @@ impl Server {
Err(_) => return Err(Error::SocketError),
};
md5_password(&mut stream, &user.username, &user.password, &salt[..])
md5_password(&mut stream, &user.name, &user.password, &salt[..])
.await?;
}
AUTHENTICATION_SUCCESSFUL => (),
SASL => {
debug!("Starting SASL authentication");
let sasl_len = (len - 8) as usize;
let mut sasl_auth = vec![0u8; sasl_len];
match stream.read_exact(&mut sasl_auth).await {
Ok(_) => (),
Err(_) => return Err(Error::SocketError),
};
let sasl_type = String::from_utf8_lossy(&sasl_auth[..sasl_len - 2]);
if sasl_type == SCRAM_SHA_256 {
debug!("Using {}", SCRAM_SHA_256);
// Generate client message.
let sasl_response = scram.message();
// SASLInitialResponse (F)
let mut res = BytesMut::new();
res.put_u8(b'p');
// length + String length + length + length of sasl response
res.put_i32(
4 // i32 size
+ SCRAM_SHA_256.len() as i32 // length of SASL version string,
+ 1 // Null terminator for the SASL version string,
+ 4 // i32 size
+ sasl_response.len() as i32, // length of SASL response
);
res.put_slice(&format!("{}\0", SCRAM_SHA_256).as_bytes()[..]);
res.put_i32(sasl_response.len() as i32);
res.put(sasl_response);
write_all(&mut stream, res).await?;
} else {
error!("Unsupported SCRAM version: {}", sasl_type);
return Err(Error::ServerError);
}
}
SASL_CONTINUE => {
trace!("Continuing SASL");
let mut sasl_data = vec![0u8; (len - 8) as usize];
match stream.read_exact(&mut sasl_data).await {
Ok(_) => (),
Err(_) => return Err(Error::SocketError),
};
let msg = BytesMut::from(&sasl_data[..]);
let sasl_response = scram.update(&msg)?;
// SASLResponse
let mut res = BytesMut::new();
res.put_u8(b'p');
res.put_i32(4 + sasl_response.len() as i32);
res.put(sasl_response);
write_all(&mut stream, res).await?;
}
SASL_FINAL => {
trace!("Final SASL");
let mut sasl_final = vec![0u8; len as usize - 8];
match stream.read_exact(&mut sasl_final).await {
Ok(_) => (),
Err(_) => return Err(Error::SocketError),
};
match scram.finish(&BytesMut::from(&sasl_final[..])) {
Ok(_) => {
debug!("SASL authentication successful");
}
Err(err) => {
debug!("SASL authentication failed");
return Err(err);
}
};
}
_ => {
error!("Unsupported authentication mechanism: {}", auth_code);
return Err(Error::ServerError);
@@ -288,7 +187,7 @@ impl Server {
// BackendKeyData
'K' => {
// The frontend must save these values if it wishes to be able to issue CancelRequest messages later.
// See: <https://www.postgresql.org/docs/12/protocol-message-formats.html>.
// See: https://www.postgresql.org/docs/12/protocol-message-formats.html
process_id = match stream.read_i32().await {
Ok(id) => id,
Err(_) => return Err(Error::SocketError),
@@ -309,9 +208,11 @@ impl Server {
Err(_) => return Err(Error::SocketError),
};
// This is the last step in the client-server connection setup,
// and indicates the server is ready for to query it.
let (read, write) = stream.into_split();
let mut server = Server {
return Ok(Server {
address: address.clone(),
read: BufReader::new(read),
write: write,
@@ -325,13 +226,7 @@ impl Server {
client_server_map: client_server_map,
connected_at: chrono::offset::Utc::now().naive_utc(),
stats: stats,
application_name: String::new(),
last_activity: SystemTime::now(),
};
server.set_name("pgcat").await?;
return Ok(server);
});
}
// We have an unexpected message from the server during this exchange.
@@ -348,7 +243,7 @@ impl Server {
/// Uses a separate connection that's not part of the connection pool.
pub async fn cancel(
host: &str,
port: u16,
port: &str,
process_id: i32,
secret_key: i32,
) -> Result<(), Error> {
@@ -373,15 +268,10 @@ impl Server {
/// Send messages to the server from the client.
pub async fn send(&mut self, messages: BytesMut) -> Result<(), Error> {
self.stats
.data_sent(messages.len(), self.process_id, self.address.id);
self.stats.data_sent(messages.len());
match write_all_half(&mut self.write, messages).await {
Ok(_) => {
// Successfully sent to server
self.last_activity = SystemTime::now();
Ok(())
}
Ok(_) => Ok(()),
Err(err) => {
error!("Terminating server because of: {:?}", err);
self.bad = true;
@@ -428,7 +318,7 @@ impl Server {
self.in_transaction = false;
}
// Some error occurred, the transaction was rolled back.
// Some error occured, the transaction was rolled back.
'E' => {
self.in_transaction = true;
}
@@ -451,7 +341,8 @@ impl Server {
// More data is available after this message, this is not the end of the reply.
self.data_available = true;
// Don't flush yet, the more we buffer, the faster this goes...up to a limit.
// Don't flush yet, the more we buffer, the faster this goes...
// up to a limit of course.
if self.buffer.len() >= 8196 {
break;
}
@@ -483,15 +374,11 @@ impl Server {
let bytes = self.buffer.clone();
// Keep track of how much data we got from the server for stats.
self.stats
.data_received(bytes.len(), self.process_id, self.address.id);
self.stats.data_received(bytes.len());
// Clear the buffer for next query.
self.buffer.clear();
// Successfully received data from server
self.last_activity = SystemTime::now();
// Pass the data back to the client.
Ok(bytes)
}
@@ -522,7 +409,7 @@ impl Server {
/// Indicate that this server connection cannot be re-used and must be discarded.
pub fn mark_bad(&mut self) {
error!("Server {:?} marked bad", self.address);
error!("Server marked bad");
self.bad = true;
}
@@ -535,7 +422,7 @@ impl Server {
self.process_id,
self.secret_key,
self.address.host.clone(),
self.address.port,
self.address.port.clone(),
),
);
}
@@ -562,14 +449,9 @@ impl Server {
/// A shorthand for `SET application_name = $1`.
#[allow(dead_code)]
pub async fn set_name(&mut self, name: &str) -> Result<(), Error> {
if self.application_name != name {
self.application_name = name.to_string();
Ok(self
.query(&format!("SET application_name = '{}'", name))
.await?)
} else {
Ok(())
}
Ok(self
.query(&format!("SET application_name = '{}'", name))
.await?)
}
/// Get the servers address.
@@ -578,15 +460,9 @@ impl Server {
self.address.clone()
}
/// Get the server's unique identifier.
pub fn process_id(&self) -> i32 {
self.process_id
}
// Get server's latest response timestamp
pub fn last_activity(&self) -> SystemTime {
self.last_activity
}
}
impl Drop for Server {
@@ -594,8 +470,7 @@ impl Drop for Server {
/// the socket is in non-blocking mode, so it may not be ready
/// for a write.
fn drop(&mut self) {
self.stats
.server_disconnecting(self.process_id(), self.address.id);
self.stats.server_disconnecting(self.process_id());
let mut bytes = BytesMut::with_capacity(4);
bytes.put_u8(b'X');
@@ -603,10 +478,9 @@ impl Drop for Server {
match self.write.try_write(&bytes) {
Ok(_) => (),
Err(_) => debug!("Dirty shutdown"),
Err(_) => (),
};
// Should not matter.
self.bad = true;
let now = chrono::offset::Utc::now().naive_utc();

View File

@@ -1,27 +1,20 @@
/// Implements various sharding functions.
use sha1::{Digest, Sha1};
/// See: <https://github.com/postgres/postgres/blob/27b77ecf9f4d5be211900eda54d8155ada50d696/src/include/catalog/partition.h#L20>.
// https://github.com/postgres/postgres/blob/27b77ecf9f4d5be211900eda54d8155ada50d696/src/include/catalog/partition.h#L20
const PARTITION_HASH_SEED: u64 = 0x7A5B22367996DCFD;
/// The sharding functions we support.
#[derive(Debug, PartialEq, Copy, Clone)]
pub enum ShardingFunction {
PgBigintHash,
Sha1,
}
/// The sharder.
pub struct Sharder {
/// Number of shards in the cluster.
shards: usize,
/// The sharding function in use.
sharding_function: ShardingFunction,
}
impl Sharder {
/// Create new instance of the sharder.
pub fn new(shards: usize, sharding_function: ShardingFunction) -> Sharder {
Sharder {
shards,
@@ -29,7 +22,6 @@ impl Sharder {
}
}
/// Compute the shard given sharding key.
pub fn shard(&self, key: i64) -> usize {
match self.sharding_function {
ShardingFunction::PgBigintHash => self.pg_bigint_hash(key),
@@ -39,7 +31,7 @@ impl Sharder {
/// Hash function used by Postgres to determine which partition
/// to put the row in when using HASH(column) partitioning.
/// Source: <https://github.com/postgres/postgres/blob/27b77ecf9f4d5be211900eda54d8155ada50d696/src/common/hashfn.c#L631>.
/// Source: https://github.com/postgres/postgres/blob/27b77ecf9f4d5be211900eda54d8155ada50d696/src/common/hashfn.c#L631
/// Supports only 1 bigint at the moment, but we can add more later.
fn pg_bigint_hash(&self, key: i64) -> usize {
let mut lohalf = key as u32;
@@ -127,7 +119,6 @@ impl Sharder {
a
}
#[inline]
fn pg_u32_hash(k: u32) -> u64 {
let mut a: u32 = 0x9e3779b9 as u32 + std::mem::size_of::<u32>() as u32 + 3923095 as u32;
let mut b = a;

View File

@@ -1,27 +1,17 @@
use arc_swap::ArcSwap;
/// Statistics and reporting.
use log::{error, info, trace};
use log::{debug, info};
use once_cell::sync::Lazy;
use parking_lot::Mutex;
use statsd::Client;
use tokio::sync::mpsc::{Receiver, Sender};
use std::collections::HashMap;
use tokio::sync::mpsc::error::TrySendError;
use tokio::sync::mpsc::{channel, Receiver, Sender};
use crate::pool::get_number_of_addresses;
use crate::config::get_config;
pub static REPORTER: Lazy<ArcSwap<Reporter>> =
Lazy::new(|| ArcSwap::from_pointee(Reporter::default()));
// Stats used in SHOW STATS
static LATEST_STATS: Lazy<Mutex<HashMap<String, i64>>> = Lazy::new(|| Mutex::new(HashMap::new()));
static STAT_PERIOD: u64 = 15000; //15 seconds
/// Latest stats updated every second; used in SHOW STATS and other admin commands.
static LATEST_STATS: Lazy<Mutex<HashMap<usize, HashMap<String, i64>>>> =
Lazy::new(|| Mutex::new(HashMap::new()));
/// Statistics period used for average calculations.
/// 15 seconds.
static STAT_PERIOD: u64 = 15000;
/// The names for the events reported
/// to the statistics collector.
#[derive(Debug, Clone, Copy)]
enum EventName {
CheckoutTime,
@@ -38,292 +28,198 @@ enum EventName {
ServerTested,
ServerLogin,
ServerDisconnecting,
UpdateStats,
UpdateAverages,
FlushStatsToStatsD,
}
/// Event data sent to the collector
/// from clients and servers.
#[derive(Debug, Clone)]
#[derive(Debug)]
pub struct Event {
/// The name of the event being reported.
name: EventName,
/// The value being reported. Meaning differs based on event name.
value: i64,
/// The client or server connection reporting the event.
process_id: i32,
/// The server the client is connected to.
address_id: usize,
process_id: Option<i32>,
}
/// The statistics reporter. An instance is given
/// to each possible source of statistics,
/// e.g. clients, servers, connection pool.
#[derive(Clone, Debug)]
pub struct Reporter {
tx: Sender<Event>,
}
impl Default for Reporter {
fn default() -> Reporter {
let (tx, _rx) = channel(5);
Reporter { tx }
}
}
impl Reporter {
/// Create a new Reporter instance.
pub fn new(tx: Sender<Event>) -> Reporter {
Reporter { tx: tx }
}
/// Send statistics to the task keeping track of stats.
fn send(&self, event: Event) {
let name = event.name;
let result = self.tx.try_send(event);
match result {
Ok(_) => trace!(
"{:?} event reported successfully, capacity: {}",
name,
self.tx.capacity()
),
Err(err) => match err {
TrySendError::Full { .. } => error!("{:?} event dropped, buffer full", name),
TrySendError::Closed { .. } => error!("{:?} event dropped, channel closed", name),
},
};
}
/// Report a query executed by a client against
/// a server identified by the `address_id`.
pub fn query(&self, process_id: i32, address_id: usize) {
pub fn query(&self) {
let event = Event {
name: EventName::Query,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: None,
};
self.send(event);
let _ = self.tx.try_send(event);
}
/// Report a transaction executed by a client against
/// a server identified by the `address_id`.
pub fn transaction(&self, process_id: i32, address_id: usize) {
pub fn transaction(&self) {
let event = Event {
name: EventName::Transaction,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: None,
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Report data sent to a server identified by `address_id`.
/// The `amount` is measured in bytes.
pub fn data_sent(&self, amount: usize, process_id: i32, address_id: usize) {
pub fn data_sent(&self, amount: usize) {
let event = Event {
name: EventName::DataSent,
value: amount as i64,
process_id: process_id,
address_id: address_id,
process_id: None,
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Report data received from a server identified by `address_id`.
/// The `amount` is measured in bytes.
pub fn data_received(&self, amount: usize, process_id: i32, address_id: usize) {
pub fn data_received(&self, amount: usize) {
let event = Event {
name: EventName::DataReceived,
value: amount as i64,
process_id: process_id,
address_id: address_id,
process_id: None,
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Time spent waiting to get a healthy connection from the pool
/// for a server identified by `address_id`.
/// Measured in milliseconds.
pub fn checkout_time(&self, ms: u128, process_id: i32, address_id: usize) {
pub fn checkout_time(&self, ms: u128) {
let event = Event {
name: EventName::CheckoutTime,
value: ms as i64,
process_id: process_id,
address_id: address_id,
process_id: None,
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a client identified by `process_id` waiting for a connection
/// to a server identified by `address_id`.
pub fn client_waiting(&self, process_id: i32, address_id: usize) {
pub fn client_waiting(&self, process_id: i32) {
let event = Event {
name: EventName::ClientWaiting,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a client identified by `process_id` is done waiting for a connection
/// to a server identified by `address_id` and is about to query the server.
pub fn client_active(&self, process_id: i32, address_id: usize) {
pub fn client_active(&self, process_id: i32) {
let event = Event {
name: EventName::ClientActive,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a client identified by `process_id` is done querying the server
/// identified by `address_id` and is no longer active.
pub fn client_idle(&self, process_id: i32, address_id: usize) {
pub fn client_idle(&self, process_id: i32) {
let event = Event {
name: EventName::ClientIdle,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a client identified by `process_id` is disconecting from the pooler.
/// The last server it was connected to is identified by `address_id`.
pub fn client_disconnecting(&self, process_id: i32, address_id: usize) {
pub fn client_disconnecting(&self, process_id: i32) {
let event = Event {
name: EventName::ClientDisconnecting,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a server connection identified by `process_id` for
/// a configured server identified by `address_id` is actively used
/// by a client.
pub fn server_active(&self, process_id: i32, address_id: usize) {
pub fn server_active(&self, process_id: i32) {
let event = Event {
name: EventName::ServerActive,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a server connection identified by `process_id` for
/// a configured server identified by `address_id` is no longer
/// actively used by a client and is now idle.
pub fn server_idle(&self, process_id: i32, address_id: usize) {
pub fn server_idle(&self, process_id: i32) {
let event = Event {
name: EventName::ServerIdle,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a server connection identified by `process_id` for
/// a configured server identified by `address_id` is attempting
/// to login.
pub fn server_login(&self, process_id: i32, address_id: usize) {
pub fn server_login(&self, process_id: i32) {
let event = Event {
name: EventName::ServerLogin,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a server connection identified by `process_id` for
/// a configured server identified by `address_id` is being
/// tested before being given to a client.
pub fn server_tested(&self, process_id: i32, address_id: usize) {
pub fn server_tested(&self, process_id: i32) {
let event = Event {
name: EventName::ServerTested,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
/// Reports a server connection identified by `process_id` is disconecting from the pooler.
/// The configured server it was connected to is identified by `address_id`.
pub fn server_disconnecting(&self, process_id: i32, address_id: usize) {
pub fn server_disconnecting(&self, process_id: i32) {
let event = Event {
name: EventName::ServerDisconnecting,
value: 1,
process_id: process_id,
address_id: address_id,
process_id: Some(process_id),
};
self.send(event)
let _ = self.tx.try_send(event);
}
}
/// The statistics collector which is receiving statistics
/// from clients, servers, and the connection pool. There is
/// only one collector (kind of like a singleton).
/// The collector can trigger events on its own, e.g.
/// it updates aggregates every second and averages every
/// 15 seconds.
pub struct Collector {
rx: Receiver<Event>,
tx: Sender<Event>,
client: Client,
}
impl Collector {
/// Create a new collector instance. There should only be one instance
/// at a time. This is ensured by mpsc which allows only one receiver.
pub fn new(rx: Receiver<Event>, tx: Sender<Event>) -> Collector {
Collector { rx, tx }
Collector {
rx,
tx,
client: Client::new(&get_config().general.statsd_address, "pgcat").unwrap(),
}
}
/// The statistics collection handler. It will collect statistics
/// for `address_id`s starting at 0 up to `addresses`.
pub async fn collect(&mut self) {
info!("Events reporter started");
let stats_template = HashMap::from([
let mut stats = HashMap::from([
("total_query_count", 0),
("total_query_time", 0),
("total_received", 0),
("total_sent", 0),
("total_xact_count", 0),
("total_sent", 0),
("total_received", 0),
("total_xact_time", 0),
("total_query_time", 0),
("total_wait_time", 0),
("avg_query_count", 0),
("avg_query_time", 0),
("avg_recv", 0),
("avg_sent", 0),
("avg_xact_count", 0),
("avg_xact_time", 0),
("avg_query_time", 0),
("avg_xact_count", 0),
("avg_sent", 0),
("avg_received", 0),
("avg_wait_time", 0),
("maxwait_us", 0),
("maxwait", 0),
@@ -336,49 +232,25 @@ impl Collector {
("sv_tested", 0),
]);
let mut stats = HashMap::new();
// Stats saved after each iteration of the flush event. Used in calculation
// of averages in the last flush period.
let mut old_stats: HashMap<usize, HashMap<String, i64>> = HashMap::new();
let mut old_stats: HashMap<String, i64> = HashMap::new();
// Track which state the client and server are at any given time.
let mut client_server_states: HashMap<usize, HashMap<i32, EventName>> = HashMap::new();
let mut client_server_states: HashMap<i32, EventName> = HashMap::new();
// Flush stats to StatsD and calculate averages every 15 seconds.
let tx = self.tx.clone();
tokio::task::spawn(async move {
let mut interval =
tokio::time::interval(tokio::time::Duration::from_millis(STAT_PERIOD / 15));
loop {
interval.tick().await;
let address_count = get_number_of_addresses();
for address_id in 0..address_count {
let _ = tx.try_send(Event {
name: EventName::UpdateStats,
value: 0,
process_id: -1,
address_id: address_id,
});
}
}
});
let tx = self.tx.clone();
tokio::task::spawn(async move {
let mut interval =
tokio::time::interval(tokio::time::Duration::from_millis(STAT_PERIOD));
loop {
interval.tick().await;
let address_count = get_number_of_addresses();
for address_id in 0..address_count {
let _ = tx.try_send(Event {
name: EventName::UpdateAverages,
value: 0,
process_id: -1,
address_id: address_id,
});
}
let _ = tx.try_send(Event {
name: EventName::FlushStatsToStatsD,
value: 0,
process_id: None,
});
}
});
@@ -392,14 +264,6 @@ impl Collector {
}
};
let stats = stats
.entry(stat.address_id)
.or_insert(stats_template.clone());
let client_server_states = client_server_states
.entry(stat.address_id)
.or_insert(HashMap::new());
let old_stats = old_stats.entry(stat.address_id).or_insert(HashMap::new());
// Some are counters, some are gauges...
match stat.name {
EventName::Query => {
@@ -449,16 +313,16 @@ impl Collector {
| EventName::ServerIdle
| EventName::ServerTested
| EventName::ServerLogin => {
client_server_states.insert(stat.process_id, stat.name);
client_server_states.insert(stat.process_id.unwrap(), stat.name);
}
EventName::ClientDisconnecting | EventName::ServerDisconnecting => {
client_server_states.remove(&stat.process_id);
client_server_states.remove(&stat.process_id.unwrap());
}
EventName::UpdateStats => {
EventName::FlushStatsToStatsD => {
// Calculate connection states
for (_, state) in client_server_states.iter() {
for (_, state) in &client_server_states {
match state {
EventName::ClientActive => {
let counter = stats.entry("cl_active").or_insert(0);
@@ -470,6 +334,11 @@ impl Collector {
*counter += 1;
}
EventName::ClientIdle => {
let counter = stats.entry("cl_idle").or_insert(0);
*counter += 1;
}
EventName::ServerIdle => {
let counter = stats.entry("sv_idle").or_insert(0);
*counter += 1;
@@ -490,20 +359,39 @@ impl Collector {
*counter += 1;
}
EventName::ClientIdle => {
let counter = stats.entry("cl_idle").or_insert(0);
*counter += 1;
}
_ => unreachable!(),
};
}
// Calculate averages
for stat in &[
"avg_query_count",
"avgxact_count",
"avg_sent",
"avg_received",
"avg_wait_time",
] {
let total_name = stat.replace("avg_", "total_");
let old_value = old_stats.entry(total_name.clone()).or_insert(0);
let new_value = stats.get(total_name.as_str()).unwrap_or(&0).to_owned();
let avg = (new_value - *old_value) / (STAT_PERIOD as i64 / 1_000); // Avg / second
stats.insert(stat, avg);
*old_value = new_value;
}
debug!("{:?}", stats);
// Update latest stats used in SHOW STATS
let mut guard = LATEST_STATS.lock();
for (key, value) in &stats {
guard.insert(key.to_string(), value.clone());
}
let mut pipeline = self.client.pipeline();
for (key, value) in stats.iter() {
let entry = guard.entry(stat.address_id).or_insert(HashMap::new());
entry.insert(key.to_string(), value.clone());
pipeline.gauge(key, *value as f64);
}
// These are re-calculated every iteration of the loop, so we don't want to add values
@@ -521,44 +409,14 @@ impl Collector {
] {
stats.insert(stat, 0);
}
}
EventName::UpdateAverages => {
// Calculate averages
for stat in &[
"avg_query_count",
"avg_query_time",
"avg_recv",
"avg_sent",
"avg_xact_time",
"avg_xact_count",
"avg_wait_time",
] {
let total_name = match stat {
&"avg_recv" => "total_received".to_string(), // Because PgBouncer is saving bytes
_ => stat.replace("avg_", "total_"),
};
let old_value = old_stats.entry(total_name.clone()).or_insert(0);
let new_value = stats.get(total_name.as_str()).unwrap_or(&0).to_owned();
let avg = (new_value - *old_value) / (STAT_PERIOD as i64 / 1_000); // Avg / second
stats.insert(stat, avg);
*old_value = new_value;
}
pipeline.send(&self.client);
}
};
}
}
}
/// Get a snapshot of statistics. Updated once a second
/// by the `Collector`.
pub fn get_stats() -> HashMap<usize, HashMap<String, i64>> {
pub fn get_stats() -> HashMap<String, i64> {
LATEST_STATS.lock().clone()
}
/// Get the statistics reporter used to update stats across the pools/clients.
pub fn get_reporter() -> Reporter {
(*(*REPORTER.load())).clone()
}

View File

@@ -1,57 +0,0 @@
// Stream wrapper.
use rustls_pemfile::{certs, rsa_private_keys};
use std::path::Path;
use std::sync::Arc;
use tokio_rustls::rustls::{self, Certificate, PrivateKey};
use tokio_rustls::TlsAcceptor;
use crate::config::get_config;
use crate::errors::Error;
// TLS
pub fn load_certs(path: &Path) -> std::io::Result<Vec<Certificate>> {
certs(&mut std::io::BufReader::new(std::fs::File::open(path)?))
.map_err(|_| std::io::Error::new(std::io::ErrorKind::InvalidInput, "invalid cert"))
.map(|mut certs| certs.drain(..).map(Certificate).collect())
}
pub fn load_keys(path: &Path) -> std::io::Result<Vec<PrivateKey>> {
rsa_private_keys(&mut std::io::BufReader::new(std::fs::File::open(path)?))
.map_err(|_| std::io::Error::new(std::io::ErrorKind::InvalidInput, "invalid key"))
.map(|mut keys| keys.drain(..).map(PrivateKey).collect())
}
pub struct Tls {
pub acceptor: TlsAcceptor,
}
impl Tls {
pub fn new() -> Result<Self, Error> {
let config = get_config();
let certs = match load_certs(&Path::new(&config.general.tls_certificate.unwrap())) {
Ok(certs) => certs,
Err(_) => return Err(Error::TlsError),
};
let mut keys = match load_keys(&Path::new(&config.general.tls_private_key.unwrap())) {
Ok(keys) => keys,
Err(_) => return Err(Error::TlsError),
};
let config = match rustls::ServerConfig::builder()
.with_safe_defaults()
.with_no_client_auth()
.with_single_cert(certs, keys.remove(0))
.map_err(|err| std::io::Error::new(std::io::ErrorKind::InvalidInput, err))
{
Ok(c) => c,
Err(_) => return Err(Error::TlsError),
};
Ok(Tls {
acceptor: TlsAcceptor::from(Arc::new(config)),
})
}
}

4
src/userlist.json Normal file
View File

@@ -0,0 +1,4 @@
{
"sven": "clear_text_password",
"sharding_user": "sharding_user"
}

57
src/userlist.rs Normal file
View File

@@ -0,0 +1,57 @@
use arc_swap::{ArcSwap, Guard};
use log::{error};
use once_cell::sync::Lazy;
use tokio::fs::File;
use tokio::io::AsyncReadExt;
use std::collections::{HashMap};
use std::sync::Arc;
use crate::errors::Error;
pub type UserList = HashMap<String, String>;
static USER_LIST: Lazy<ArcSwap<UserList>> = Lazy::new(|| ArcSwap::from_pointee(HashMap::new()));
pub fn get_user_list() -> Guard<Arc<UserList>> {
USER_LIST.load()
}
/// Parse the user list.
pub async fn parse(path: &str) -> Result<(), Error> {
let mut contents = String::new();
let mut file = match File::open(path).await {
Ok(file) => file,
Err(err) => {
error!("Could not open '{}': {}", path, err.to_string());
return Err(Error::BadConfig);
}
};
match file.read_to_string(&mut contents).await {
Ok(_) => (),
Err(err) => {
error!("Could not read config file: {}", err.to_string());
return Err(Error::BadConfig);
}
};
let map: HashMap<String, String> = serde_json::from_str(&contents).expect("JSON was not well-formatted");
USER_LIST.store(Arc::new(map.clone()));
Ok(())
}
#[cfg(test)]
mod test {
use super::*;
#[tokio::test]
async fn test_config() {
parse("userlist.json").await.unwrap();
assert_eq!(get_user_list()["sven"], "clear_text_password");
assert_eq!(get_user_list()["sharding_user"], "sharding_user");
}
}

View File

@@ -1,39 +0,0 @@
-- \setrandom aid 1 :naccounts
\set aid random(1, 100000)
-- \setrandom bid 1 :nbranches
\set bid random(1, 100000)
-- \setrandom tid 1 :ntellers
\set tid random(1, 100000)
-- \setrandom delta -5000 5000
\set delta random(-5000,5000)
\set shard random(0, 2)
SET SHARD TO :shard;
SET SERVER ROLE TO 'auto';
BEGIN;
UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
END;
SET SHARDING KEY TO :aid;
-- Read load balancing
SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
SET SERVER ROLE TO 'replica';
-- Read load balancing
SELECT abalance FROM pgbench_accounts WHERE aid = :aid;

View File

@@ -1,2 +1 @@
psycopg2==2.9.3
psutil==5.9.1

View File

@@ -1,252 +1,11 @@
from typing import Tuple
import psycopg2
import psutil
import os
import signal
import time
SHUTDOWN_TIMEOUT = 5
conn = psycopg2.connect("postgres://random:password@127.0.0.1:6432/db")
cur = conn.cursor()
PGCAT_HOST = "127.0.0.1"
PGCAT_PORT = "6432"
cur.execute("SELECT 1");
res = cur.fetchall()
print(res)
def pgcat_start():
pg_cat_send_signal(signal.SIGTERM)
os.system("./target/debug/pgcat .circleci/pgcat.toml &")
time.sleep(2)
def pg_cat_send_signal(signal: signal.Signals):
for proc in psutil.process_iter(["pid", "name"]):
if "pgcat" == proc.name():
os.kill(proc.pid, signal)
if signal == signal.SIGTERM:
# Returns 0 if pgcat process exists
time.sleep(2)
if not os.system('pgrep pgcat'):
raise Exception("pgcat not closed after SIGTERM")
def connect_db(
autocommit: bool = True,
admin: bool = False,
) -> Tuple[psycopg2.extensions.connection, psycopg2.extensions.cursor]:
if admin:
user = "admin_user"
password = "admin_pass"
db = "pgcat"
else:
user = "sharding_user"
password = "sharding_user"
db = "sharded_db"
conn = psycopg2.connect(
f"postgres://{user}:{password}@{PGCAT_HOST}:{PGCAT_PORT}/{db}?application_name=testing_pgcat",
connect_timeout=2,
)
conn.autocommit = autocommit
cur = conn.cursor()
return (conn, cur)
def cleanup_conn(conn: psycopg2.extensions.connection, cur: psycopg2.extensions.cursor):
cur.close()
conn.close()
def test_normal_db_access():
conn, cur = connect_db(autocommit=False)
cur.execute("SELECT 1")
res = cur.fetchall()
print(res)
cleanup_conn(conn, cur)
def test_admin_db_access():
conn, cur = connect_db(admin=True)
cur.execute("SHOW POOLS")
res = cur.fetchall()
print(res)
cleanup_conn(conn, cur)
def test_shutdown_logic():
# - - - - - - - - - - - - - - - - - -
# NO ACTIVE QUERIES SIGINT HANDLING
# Start pgcat
pgcat_start()
# Create client connection and send query (not in transaction)
conn, cur = connect_db()
cur.execute("BEGIN;")
cur.execute("SELECT 1;")
cur.execute("COMMIT;")
# Send sigint to pgcat
pg_cat_send_signal(signal.SIGINT)
time.sleep(1)
# Check that any new queries fail after sigint since server should close with no active transactions
try:
cur.execute("SELECT 1;")
except psycopg2.OperationalError as e:
pass
else:
# Fail if query execution succeeded
raise Exception("Server not closed after sigint")
cleanup_conn(conn, cur)
pg_cat_send_signal(signal.SIGTERM)
# - - - - - - - - - - - - - - - - - -
# HANDLE TRANSACTION WITH SIGINT
# Start pgcat
pgcat_start()
# Create client connection and begin transaction
conn, cur = connect_db()
cur.execute("BEGIN;")
cur.execute("SELECT 1;")
# Send sigint to pgcat while still in transaction
pg_cat_send_signal(signal.SIGINT)
time.sleep(1)
# Check that any new queries succeed after sigint since server should still allow transaction to complete
try:
cur.execute("SELECT 1;")
except psycopg2.OperationalError as e:
# Fail if query fails since server closed
raise Exception("Server closed while in transaction", e.pgerror)
cleanup_conn(conn, cur)
pg_cat_send_signal(signal.SIGTERM)
# - - - - - - - - - - - - - - - - - -
# NO NEW NON-ADMIN CONNECTIONS DURING SHUTDOWN
# Start pgcat
pgcat_start()
# Create client connection and begin transaction
transaction_conn, transaction_cur = connect_db()
transaction_cur.execute("BEGIN;")
transaction_cur.execute("SELECT 1;")
# Send sigint to pgcat while still in transaction
pg_cat_send_signal(signal.SIGINT)
time.sleep(1)
start = time.perf_counter()
try:
conn, cur = connect_db()
cur.execute("SELECT 1;")
cleanup_conn(conn, cur)
except psycopg2.OperationalError as e:
time_taken = time.perf_counter() - start
if time_taken > 0.1:
raise Exception(
"Failed to reject connection within 0.1 seconds, got", time_taken, "seconds")
pass
else:
raise Exception("Able connect to database during shutdown")
cleanup_conn(transaction_conn, transaction_cur)
pg_cat_send_signal(signal.SIGTERM)
# - - - - - - - - - - - - - - - - - -
# ALLOW NEW ADMIN CONNECTIONS DURING SHUTDOWN
# Start pgcat
pgcat_start()
# Create client connection and begin transaction
transaction_conn, transaction_cur = connect_db()
transaction_cur.execute("BEGIN;")
transaction_cur.execute("SELECT 1;")
# Send sigint to pgcat while still in transaction
pg_cat_send_signal(signal.SIGINT)
time.sleep(1)
try:
conn, cur = connect_db(admin=True)
cur.execute("SHOW DATABASES;")
cleanup_conn(conn, cur)
except psycopg2.OperationalError as e:
raise Exception(e)
cleanup_conn(transaction_conn, transaction_cur)
pg_cat_send_signal(signal.SIGTERM)
# - - - - - - - - - - - - - - - - - -
# ADMIN CONNECTIONS CONTINUING TO WORK AFTER SHUTDOWN
# Start pgcat
pgcat_start()
# Create client connection and begin transaction
transaction_conn, transaction_cur = connect_db()
transaction_cur.execute("BEGIN;")
transaction_cur.execute("SELECT 1;")
admin_conn, admin_cur = connect_db(admin=True)
admin_cur.execute("SHOW DATABASES;")
# Send sigint to pgcat while still in transaction
pg_cat_send_signal(signal.SIGINT)
time.sleep(1)
try:
admin_cur.execute("SHOW DATABASES;")
except psycopg2.OperationalError as e:
raise Exception("Could not execute admin command:", e)
cleanup_conn(transaction_conn, transaction_cur)
cleanup_conn(admin_conn, admin_cur)
pg_cat_send_signal(signal.SIGTERM)
# - - - - - - - - - - - - - - - - - -
# HANDLE SHUTDOWN TIMEOUT WITH SIGINT
# Start pgcat
pgcat_start()
# Create client connection and begin transaction, which should prevent server shutdown unless shutdown timeout is reached
conn, cur = connect_db()
cur.execute("BEGIN;")
cur.execute("SELECT 1;")
# Send sigint to pgcat while still in transaction
pg_cat_send_signal(signal.SIGINT)
# pgcat shutdown timeout is set to SHUTDOWN_TIMEOUT seconds, so we sleep for SHUTDOWN_TIMEOUT + 1 seconds
time.sleep(SHUTDOWN_TIMEOUT + 1)
# Check that any new queries succeed after sigint since server should still allow transaction to complete
try:
cur.execute("SELECT 1;")
except psycopg2.OperationalError as e:
pass
else:
# Fail if query execution succeeded
raise Exception("Server not closed after sigint and expected timeout")
cleanup_conn(conn, cur)
pg_cat_send_signal(signal.SIGTERM)
# - - - - - - - - - - - - - - - - - -
test_normal_db_access()
test_admin_db_access()
test_shutdown_logic()
# conn.commit()

View File

@@ -1,2 +1 @@
3.0.0
2.7.1

View File

@@ -2,5 +2,3 @@ source "https://rubygems.org"
gem "pg"
gem "activerecord"
gem "rubocop"
gem "toml", "~> 0.3.0"

View File

@@ -1,56 +1,30 @@
GEM
remote: https://rubygems.org/
specs:
activemodel (7.0.3.1)
activesupport (= 7.0.3.1)
activerecord (7.0.3.1)
activemodel (= 7.0.3.1)
activesupport (= 7.0.3.1)
activesupport (7.0.3.1)
activemodel (7.0.2.2)
activesupport (= 7.0.2.2)
activerecord (7.0.2.2)
activemodel (= 7.0.2.2)
activesupport (= 7.0.2.2)
activesupport (7.0.2.2)
concurrent-ruby (~> 1.0, >= 1.0.2)
i18n (>= 1.6, < 2)
minitest (>= 5.1)
tzinfo (~> 2.0)
ast (2.4.2)
concurrent-ruby (1.1.10)
i18n (1.11.0)
concurrent-ruby (1.1.9)
i18n (1.10.0)
concurrent-ruby (~> 1.0)
minitest (5.16.2)
parallel (1.22.1)
parser (3.1.2.0)
ast (~> 2.4.1)
parslet (2.0.0)
minitest (5.15.0)
pg (1.3.2)
rainbow (3.1.1)
regexp_parser (2.3.1)
rexml (3.2.5)
rubocop (1.29.0)
parallel (~> 1.10)
parser (>= 3.1.0.0)
rainbow (>= 2.2.2, < 4.0)
regexp_parser (>= 1.8, < 3.0)
rexml (>= 3.2.5, < 4.0)
rubocop-ast (>= 1.17.0, < 2.0)
ruby-progressbar (~> 1.7)
unicode-display_width (>= 1.4.0, < 3.0)
rubocop-ast (1.17.0)
parser (>= 3.1.1.0)
ruby-progressbar (1.11.0)
toml (0.3.0)
parslet (>= 1.8.0, < 3.0.0)
tzinfo (2.0.4)
concurrent-ruby (~> 1.0)
unicode-display_width (2.1.0)
PLATFORMS
arm64-darwin-21
x86_64-linux
DEPENDENCIES
activerecord
pg
rubocop
toml (~> 0.3.0)
BUNDLED WITH
2.3.7

View File

@@ -1,93 +1,6 @@
# frozen_string_literal: true
require 'active_record'
require 'pg'
require 'toml'
$stdout.sync = true
$stderr.sync = true
class ConfigEditor
def initialize
@original_config_text = File.read('../../.circleci/pgcat.toml')
text_to_load = @original_config_text.gsub("5432", "\"5432\"")
@original_configs = TOML.load(text_to_load)
end
def original_configs
TOML.load(TOML::Generator.new(@original_configs).body)
end
def with_modified_configs(new_configs)
text_to_write = TOML::Generator.new(new_configs).body
text_to_write = text_to_write.gsub("\"5432\"", "5432")
File.write('../../.circleci/pgcat.toml', text_to_write)
yield
ensure
File.write('../../.circleci/pgcat.toml', @original_config_text)
end
end
def with_captured_stdout_stderr
sout = STDOUT.clone
serr = STDERR.clone
STDOUT.reopen("/tmp/out.txt", "w+")
STDERR.reopen("/tmp/err.txt", "w+")
STDOUT.sync = true
STDERR.sync = true
yield
return File.read('/tmp/out.txt'), File.read('/tmp/err.txt')
ensure
STDOUT.reopen(sout)
STDERR.reopen(serr)
end
def test_extended_protocol_pooler_errors
admin_conn = PG::connect("postgres://admin_user:admin_pass@127.0.0.1:6432/pgcat")
conf_editor = ConfigEditor.new
new_configs = conf_editor.original_configs
# shorter timeouts
new_configs["general"]["connect_timeout"] = 500
new_configs["general"]["ban_time"] = 1
new_configs["general"]["shutdown_timeout"] = 1
new_configs["pools"]["sharded_db"]["users"]["0"]["pool_size"] = 1
new_configs["pools"]["sharded_db"]["users"]["1"]["pool_size"] = 1
conf_editor.with_modified_configs(new_configs) { admin_conn.async_exec("RELOAD") }
conn_str = "postgres://sharding_user:sharding_user@127.0.0.1:6432/sharded_db"
10.times do
Thread.new do
conn = PG::connect(conn_str)
conn.async_exec("SELECT pg_sleep(5)") rescue PG::SystemError
ensure
conn&.close
end
end
sleep(0.5)
conn_under_test = PG::connect(conn_str)
stdout, stderr = with_captured_stdout_stderr do
5.times do |i|
conn_under_test.async_exec("SELECT 1") rescue PG::SystemError
conn_under_test.exec_params("SELECT #{i} + $1", [i]) rescue PG::SystemError
sleep 1
end
end
raise StandardError, "Libpq got unexpected messages while idle" if stderr.include?("arrived from server while idle")
puts "Pool checkout errors not breaking clients passed"
ensure
sleep 1
admin_conn.async_exec("RELOAD") # Reset state
conn_under_test&.close
end
test_extended_protocol_pooler_errors
# Uncomment these two to see all queries.
# ActiveRecord.verbose_query_logs = true
@@ -99,8 +12,7 @@ ActiveRecord::Base.establish_connection(
port: 6432,
username: 'sharding_user',
password: 'sharding_user',
database: 'sharded_db',
application_name: 'testing_pgcat',
database: 'rails_dev',
prepared_statements: false, # Transaction mode
advisory_locks: false # Same
)
@@ -198,68 +110,3 @@ begin
rescue ActiveRecord::StatementInvalid
puts 'OK'
end
# Test evil clients
def poorly_behaved_client
conn = PG::connect("postgres://sharding_user:sharding_user@127.0.0.1:6432/sharded_db?application_name=testing_pgcat")
conn.async_exec 'BEGIN'
conn.async_exec 'SELECT 1'
conn.close
puts 'Bad client ok'
end
25.times do
poorly_behaved_client
end
def test_server_parameters
server_conn = PG::connect("postgres://sharding_user:sharding_user@127.0.0.1:6432/sharded_db?application_name=testing_pgcat")
raise StandardError, "Bad server version" if server_conn.server_version == 0
server_conn.close
admin_conn = PG::connect("postgres://admin_user:admin_pass@127.0.0.1:6432/pgcat")
raise StandardError, "Bad server version" if admin_conn.server_version == 0
admin_conn.close
puts 'Server parameters ok'
end
def test_reload_pool_recycling
admin_conn = PG::connect("postgres://admin_user:admin_pass@127.0.0.1:6432/pgcat")
server_conn = PG::connect("postgres://sharding_user:sharding_user@127.0.0.1:6432/sharded_db?application_name=testing_pgcat")
server_conn.async_exec("BEGIN")
conf_editor = ConfigEditor.new
new_configs = conf_editor.original_configs
# swap shards
new_configs["pools"]["sharded_db"]["shards"]["0"]["database"] = "shard1"
new_configs["pools"]["sharded_db"]["shards"]["1"]["database"] = "shard0"
raise StandardError if server_conn.async_exec("SELECT current_database();")[0]["current_database"] != 'shard0'
conf_editor.with_modified_configs(new_configs) { admin_conn.async_exec("RELOAD") }
raise StandardError if server_conn.async_exec("SELECT current_database();")[0]["current_database"] != 'shard0'
server_conn.async_exec("COMMIT;")
# Transaction finished, client should get new configs
raise StandardError if server_conn.async_exec("SELECT current_database();")[0]["current_database"] != 'shard1'
server_conn.close()
# New connection should get new configs
server_conn = PG::connect("postgres://sharding_user:sharding_user@127.0.0.1:6432/sharded_db?application_name=testing_pgcat")
raise StandardError if server_conn.async_exec("SELECT current_database();")[0]["current_database"] != 'shard1'
ensure
admin_conn.async_exec("RELOAD") # Go back to old state
admin_conn.close
server_conn.close
puts "Pool Recycling okay!"
end
test_reload_pool_recycling

View File

@@ -1,12 +1,11 @@
DROP DATABASE IF EXISTS shard0;
DROP DATABASE IF EXISTS shard1;
DROP DATABASE IF EXISTS shard2;
DROP DATABASE IF EXISTS some_db;
CREATE DATABASE shard0;
CREATE DATABASE shard1;
CREATE DATABASE shard2;
CREATE DATABASE some_db;
\c shard0
@@ -42,51 +41,21 @@ CREATE TABLE data (
CREATE TABLE data_shard_2 PARTITION OF data FOR VALUES WITH (MODULUS 3, REMAINDER 2);
\c some_db
DROP TABLE IF EXISTS data CASCADE;
CREATE TABLE data (
id BIGINT,
value VARCHAR
);
DROP ROLE IF EXISTS sharding_user;
DROP ROLE IF EXISTS other_user;
DROP ROLE IF EXISTS simple_user;
CREATE ROLE sharding_user ENCRYPTED PASSWORD 'sharding_user' LOGIN;
CREATE ROLE other_user ENCRYPTED PASSWORD 'other_user' LOGIN;
CREATE ROLE simple_user ENCRYPTED PASSWORD 'simple_user' LOGIN;
GRANT CONNECT ON DATABASE shard0 TO sharding_user;
GRANT CONNECT ON DATABASE shard1 TO sharding_user;
GRANT CONNECT ON DATABASE shard2 TO sharding_user;
GRANT CONNECT ON DATABASE shard0 TO other_user;
GRANT CONNECT ON DATABASE shard1 TO other_user;
GRANT CONNECT ON DATABASE shard2 TO other_user;
GRANT CONNECT ON DATABASE some_db TO simple_user;
GRANT CONNECT ON DATABASE shard0 TO sharding_user;
GRANT CONNECT ON DATABASE shard1 TO sharding_user;
GRANT CONNECT ON DATABASE shard2 TO sharding_user;
\c shard0
GRANT ALL ON SCHEMA public TO sharding_user;
GRANT ALL ON TABLE data TO sharding_user;
GRANT ALL ON SCHEMA public TO other_user;
GRANT ALL ON TABLE data TO other_user;
\c shard1
GRANT ALL ON SCHEMA public TO sharding_user;
GRANT ALL ON TABLE data TO sharding_user;
GRANT ALL ON SCHEMA public TO other_user;
GRANT ALL ON TABLE data TO other_user;
\c shard2
GRANT ALL ON SCHEMA public TO sharding_user;
GRANT ALL ON TABLE data TO sharding_user;
GRANT ALL ON SCHEMA public TO other_user;
GRANT ALL ON TABLE data TO other_user;
\c some_db
GRANT ALL ON SCHEMA public TO simple_user;
GRANT ALL ON TABLE data TO simple_user;

View File

@@ -151,12 +151,3 @@ SELECT 1;
set server role to 'replica';
SeT SeRver Role TO 'PrImARY';
select 1;
SET PRIMARY READS TO 'on';
SELECT 1;
SET PRIMARY READS TO 'off';
SELECT 1;
SET PRIMARY READS TO 'default';
SELECT 1;