v3.0.2

Not yet tagged, pending a couple of tests

CONTRIBUTING.md

@@ -0,0 +1,29 @@
+License and Contributions
+=========================
+
+`repmgr` is licensed under the GPL v3.  All of its code and documentation is
+Copyright 2010-2015, 2ndQuadrant Limited.  See the files COPYRIGHT and LICENSE for
+details.
+
+The development of repmgr has primarily been sponsored by 2ndQuadrant customers.
+
+Additional work has been sponsored by the 4CaaST project for cloud computing,
+which has received funding from the European Union's Seventh Framework Programme
+(FP7/2007-2013) under grant agreement 258862.
+
+Contributions to `repmgr` are welcome, and will be listed in the file `CREDITS`.
+2ndQuadrant Limited requires that any contributions provide a copyright
+assignment and a disclaimer of any work-for-hire ownership claims from the
+employer of the developer.  This lets us make sure that all of the repmgr
+distribution remains free code.  Please contact info@2ndQuadrant.com for a
+copy of the relevant Copyright Assignment Form.
+
+Code style
+----------
+
+Code in repmgr is formatted to a consistent style using the following command:
+
+    astyle --style=ansi --indent=tab --suffix=none *.c *.h
+
+Contributors should reformat their code similarly before submitting code to
+the project, in order to minimize merge conflicts with other work.

FAILOVER.rst

@@ -203,6 +203,12 @@ repmgr will also ask for the superuser password on the witness database so
 it can reconnect when needed (the command line option --initdb-no-pwprompt
 will set up a password-less superuser).
 
+By default the witness server will listen on port 5499; this value can be
+overridden by explicitly providing the port number in the conninfo string
+in repmgr.conf. (Note that it is also possible to specify the port number
+with the -l/--local-port option, however this option is now deprecated and
+will be overridden by a port setting in the conninfo string).
+
 Start the repmgrd daemons
 -------------------------
 

HISTORY

@@ -1,3 +1,19 @@
+3.0.2   2015-09-
+        Improve handling of --help/--version options; and improve help output (Ian)
+        Improve handling of situation where logfile can't be opened (Ian)
+        Always pass -D/--pgdata option to pg_basebackup (Ian)
+        Bugfix: standby clone --force does not empty pg_xlog (Gianni)
+        Bugfix: autofailover with reconnect_attempts > 1 (Gianni)
+        Bugfix: ignore comments after values (soxwellfb)
+        Bugfix: handle string values in 'node' parameter correctly (Gregory Duchatelet)
+        Allow repmgr to be compiled with a newer libpq (Marco)
+        Bugfix: call update_node_record_set_upstream() for STANDBY FOLLOW (Tomas)
+        Update `repmgr --help` output (per Github report from renard)
+        Update tablespace remapping in --rsync-only mode for 9.5 and later (Ian)
+        Deprecate `-l/--local-port` option - the port can be extracted
+          from the conninfo string in repmgr.conf (Ian)
+        Add STANDBY UNREGISTE (Vik Fearing)
+
 3.0.1   2015-04-16
         Prevent repmgrd from looping infinitely if node was not registered (Ian)
         When promoting a standby, have repmgr (not repmgrd) handle metadata updates (Ian)

PACKAGES.md

@@ -4,10 +4,10 @@ Packaging
 Notes on RedHat Linux, Fedora, and CentOS Builds
 ------------------------------------------------
 
-The RPM packages of PostgreSQL put ``pg_config`` into the ``postgresql-devel``
+The RPM packages of PostgreSQL put `pg_config` into the `postgresql-devel`
 package, not the main server one.  And if you have a RPM install of PostgreSQL
 9.0, the entire PostgreSQL binary directory will not be in your PATH by default
-either.  Individual utilities are made available via the ``alternatives``
+either.  Individual utilities are made available via the `alternatives`
 mechanism, but not all commands will be wrapped that way.  The files installed
 by repmgr will certainly not be in the default PATH for the postgres user
 on such a system.  They will instead be in /usr/pgsql-9.0/bin/ on this
@@ -15,31 +15,33 @@ type of system.
 
 When building repmgr against a RPM packaged build, you may discover that some
 development packages are needed as well.  The following build errors can
-occur::
+occur:
 
     /usr/bin/ld: cannot find -lxslt
     /usr/bin/ld: cannot find -lpam
 
-Install the following packages to correct those::
+Install the following packages to correct those:
+
 
     yum install libxslt-devel
     yum install pam-devel
 
 If building repmgr as a regular user, then doing the install into the system
-directories using sudo, the syntax is hard.  ``pg_config`` won't be in root's
+directories using sudo, the syntax is hard.  `pg_config` won't be in root's
-path either.  The following recipe should work::
+path either.  The following recipe should work:
 
     sudo PATH="/usr/pgsql-9.0/bin:$PATH" make USE_PGXS=1 install
 
+
 Issues with 32 and 64 bit RPMs
 ------------------------------
 
-If when building, you receive a series of errors of this form::
+If when building, you receive a series of errors of this form:
 
   /usr/bin/ld: skipping incompatible /usr/pgsql-9.0/lib/libpq.so when searching for -lpq
 
 This is likely because you have both the 32 and 64 bit versions of the
-``postgresql90-devel`` package installed.  You can check that like this::
+`postgresql90-devel` package installed.  You can check that like this:
 
     rpm -qa --queryformat '%{NAME}\t%{ARCH}\n'  | grep postgresql90-devel
 
@@ -47,7 +49,8 @@ And if two packages appear, one for i386 and one for x86_64, that's not supposed
 to be allowed.
 
 This can happen when using the PGDG repo to install that package;
-here is an example sessions demonstrating the problem case appearing::
+here is an example sessions demonstrating the problem case appearing:
+
 
     # yum install postgresql-devel
     ..
@@ -67,20 +70,21 @@ here is an example sessions demonstrating the problem case appearing::
      postgresql90-devel    i386      9.0.2-2PGDG.rhel5    pgdg90        1.5 M
      postgresql90-devel    x86_64    9.0.2-2PGDG.rhel5    pgdg90        1.6 M
 
+
 Note how both the i386 and x86_64 platform architectures are selected for
 installation.  Your main PostgreSQL package will only be compatible with one of
 those, and if the repmgr build finds the wrong postgresql90-devel these
 "skipping incompatible" messages appear.
 
 In this case, you can temporarily remove both packages, then just install the
-correct one for your architecture.  Example::
+correct one for your architecture.  Example:
 
     rpm -e postgresql90-devel --allmatches
     yum install postgresql90-devel-9.0.2-2PGDG.rhel5.x86_64
 
 Instead just deleting the package from the wrong platform might not leave behind
 the correct files, due to the way in which these accidentally happen to interact.
-If you already tried to build repmgr before doing this, you'll need to do::
+If you already tried to build repmgr before doing this, you'll need to do:
 
     make USE_PGXS=1 clean
 
@@ -89,17 +93,17 @@ to get rid of leftover files from the wrong architecture.
 Notes on Ubuntu, Debian or other Debian-based Builds
 ----------------------------------------------------
 
-The Debian packages of PostgreSQL put ``pg_config`` into the development package
+The Debian packages of PostgreSQL put `pg_config` into the development package
-called ``postgresql-server-dev-$version``.
+called `postgresql-server-dev-$version`.
 
 When building repmgr against a Debian packages build, you may discover that some
 development packages are needed as well. You will need the following development
-packages installed::
+packages installed:
 
     sudo apt-get install libxslt-dev libxml2-dev libpam-dev libedit-dev
 
 If you're using Debian packages for PostgreSQL and are building repmgr with the
-USE_PGXS option you also need to install the corresponding development package::
+USE_PGXS option you also need to install the corresponding development package:
 
     sudo apt-get install postgresql-server-dev-9.0
 
@@ -110,12 +114,12 @@ multiple installed versions of PostgreSQL on the same system through a wrapper
 called pg_wrapper and repmgr is not (yet) known to this wrapper.
 
 You can solve this in many different ways, the most Debian like is to make an
-alternate for repmgr and repmgrd::
+alternate for repmgr and repmgrd:
 
     sudo update-alternatives --install /usr/bin/repmgr repmgr /usr/lib/postgresql/9.0/bin/repmgr 10
     sudo update-alternatives --install /usr/bin/repmgrd repmgrd /usr/lib/postgresql/9.0/bin/repmgrd 10
 
-You can also make a deb package of repmgr using::
+You can also make a deb package of repmgr using:
 
     make USE_PGXS=1 deb
 

QUICKSTART.md

@@ -21,7 +21,8 @@ Master setup
 	CREATE DATABASE repmgr_db OWNER repmgr_usr;
 	```
 
-  - configure `postgresql.conf` for replication (see above)
+  - configure `postgresql.conf` for replication (see README.md for sample
+    settings)
 
   - update `pg_hba.conf`, e.g.:
 
@@ -111,7 +112,7 @@ created in the `repl_nodes` table should look something like this:
 
         repmgr_db=# SELECT * from repmgr_test.repl_nodes;
          id |  type   | upstream_node_id | cluster | name  |                      conninfo                      | slot_name | priority | active
-        ----+---------+------------------+---------+-------+-------------------------------------------------+-----------+----------+--------
+        ----+---------+------------------+---------+-------+----------------------------------------------------+-----------+----------+--------
-          1 | primary |                  | test    | node1 | host=localhost user=repmgr_usr dbname=repmgr_db |           |        0 | t
+          1 | primary |                  | test    | node1 | host=repmgr_node1 user=repmgr_usr dbname=repmgr_db |           |        0 | t
-          2 | standby |                1 | test    | node2 | host=localhost user=repmgr_usr dbname=repmgr_db |           |        0 | t
+          2 | standby |                1 | test    | node2 | host=repmgr_node2 user=repmgr_usr dbname=repmgr_db |           |        0 | t
         (2 rows)

README.md

@@ -7,7 +7,7 @@ hot-standby capabilities with tools to set up standby servers, monitor
 replication, and perform administrative tasks such as failover or manual
 switchover operations.
 
-This document covers `repmgr 3`, which supports PostgreSQL 9.4 and 9.3.
+This document covers `repmgr 3`, which supports PostgreSQL 9.3 and later.
 This version can use `pg_basebackup` to clone standby servers, supports
 replication slots and cascading replication, doesn't require a restart
 after promotion, and has many usability improvements.
@@ -53,7 +53,7 @@ on any UNIX-like system which PostgreSQL itself supports.
 
 All nodes must be running the same major version of PostgreSQL, and we
 recommend that they also run the same minor version. This version of
-`repmgr` (v3) supports PostgreSQL 9.3 and 9.4.
+`repmgr` (v3) supports PostgreSQL 9.3 and later.
 
 Earlier versions of `repmgr` needed password-less SSH access between
 nodes in order to clone standby servers using `rsync`. `repmgr 3` can
@@ -98,8 +98,8 @@ for details.
 
 ### PostgreSQL configuration
 
-The primary server needs to be configured for replication with the
+The primary server needs to be configured for replication with settings
-following settings in `postgresql.conf`:
+like the following in `postgresql.conf`:
 
     # Allow read-only queries on standby servers. The number of WAL
     # senders should be larger than the number of standby servers.
@@ -121,13 +121,18 @@ following settings in `postgresql.conf`:
     archive_mode = on
     archive_command = 'cd .'
 
-    # You can also set additional replication parameters here, such as
+    # If you plan to use repmgrd, ensure that shared_preload_libraries
-    # hot_standby_feedback or synchronous_standby_names.
+    # is configured to load 'repmgr_funcs'
+
+    shared_preload_libraries = 'repmgr_funcs'
 
 PostgreSQL 9.4 makes it possible to use replication slots, which means
-the value of wal_keep_segments need no longer be set. With 9.3, `repmgr`
+the value of `wal_keep_segments` need no longer be set. See section
-expects it to be set to at least 5000 (= 80GB of WAL) by default, though
+"Replication slots" below for more details.
-this can be overriden with the `-w N` argument.
+
+With PostgreSQL 9.3, `repmgr` expects `wal_keep_segments` to be set to
+at least 5000 (= 80GB of WAL) by default, though this can be overriden
+with the `-w N` argument.
 
 A dedicated PostgreSQL superuser account and a database in which to
 store monitoring and replication data are required. Create them by
@@ -223,7 +228,7 @@ The node can then be restarted.
 The node will then need to be re-registered with `repmgr`; again
 the `--force` option is required to update the existing record:
 
-     repmgr -f /etc/repmgr/repmgr.conf
+     repmgr -f /etc/repmgr/repmgr.conf \
        --force \
        standby register
 
@@ -345,6 +350,7 @@ Following event types currently exist:
 
     master_register
     standby_register
+    standby_unregister
     standby_clone
     standby_promote
     witness_create
@@ -398,6 +404,18 @@ stored in the `repl_nodes` table.
 Note that `repmgr` will fail with an error if this option is specified when
 working with PostgreSQL 9.3.
 
+Be aware that when initially cloning a standby, you will need to ensure
+that all required WAL files remain available while the cloning is taking
+place. If using the default `pg_basebackup` method, we recommend setting
+`pg_basebackup`'s `--xlog-method` parameter to `stream` like this:
+
+    pg_basebackup_options='--xlog-method=stream'
+
+See the `pg_basebackup` documentation [*] for details. Otherwise you'll need
+to set `wal_keep_segments` to an appropriately high value.
+
+[*] http://www.postgresql.org/docs/current/static/app-pgbasebackup.html
+
 Further reading:
  * http://www.postgresql.org/docs/current/interactive/warm-standby.html#STREAMING-REPLICATION-SLOTS
  * http://blog.2ndquadrant.com/postgresql-9-4-slots/
@@ -435,12 +453,19 @@ its port if is different from the default one.
     Registers a master in a cluster. This command needs to be executed before any
     standby nodes are registered.
 
+    `primary register` can be used as an alias for `master register`.
+
 * `standby register`
 
     Registers a standby with `repmgr`. This command needs to be executed to enable
     promote/follow operations and to allow `repmgrd` to work with the node.
     An existing standby can be registered using this command.
 
+* `standby unregister`
+
+    Unregisters a standby with `repmgr`. This command does not affect the actual
+    replication.
+
 * `standby clone [node to be cloned]`
 
     Clones a new standby node from the data directory of the master (or

RHEL/repmgrd.sysconfig

@@ -12,7 +12,7 @@ REPMGRD_ENABLED=no
 #REPMGRD_USER=postgres
 
 # repmgrd binary
-#REPMGRD_BIN=/usr/bin/repmgr
+#REPMGRD_BIN=/usr/bin/repmgrd
 
 # pid file
 #REPMGRD_PIDFILE=/var/lib/pgsql/repmgr/repmgrd.pid

SSH-RSYNC.md

@@ -1,12 +1,13 @@
 Set up trusted copy between postgres accounts
 ---------------------------------------------
 
-If you need to use rsync to clone standby servers, the postgres account
+If you need to use `rsync` to clone standby servers, the `postgres` account
-on your master and standby servers must be each able to access the other
+on your primary and standby servers must be each able to access the other
 using SSH without a password.
 
-First generate a ssh key, using an empty passphrase, and copy the resulting
+First generate an ssh key, using an empty passphrase, and copy the resulting
-keys and a maching authorization file to a privledged user on the other system::
+keys and a matching authorization file to a privileged user account on the other
+system:
 
     [postgres@node1]$ ssh-keygen -t rsa
     Generating public/private rsa key pair.
@@ -22,8 +23,8 @@ keys and a maching authorization file to a privledged user on the other system::
     [postgres@node1]$ cd ~/.ssh
     [postgres@node1]$ scp id_rsa.pub id_rsa authorized_keys user@node2:
 
-Login as a user on the other system, and install the files into the postgres
+Login as a user on the other system, and install the files into the `postgres`
-user's account::
+user's account:
 
     [user@node2 ~]$ sudo chown postgres.postgres authorized_keys id_rsa.pub id_rsa
     [user@node2 ~]$ sudo mkdir -p ~postgres/.ssh

TODO

@@ -10,6 +10,10 @@ Known issues in repmgr
 Planned feature improvements
 ============================
 
+* Use 'primary' instead of 'master' in documentation and log output
+  for consistency with PostgreSQL documentation. See also commit
+  870b0a53b627eeb9aca1fc14cbafe25b5beafe12.
+
 * A better check which standby did receive most of the data
 
 * Make the fact that a standby may be delayed a factor in the voting
@@ -22,6 +26,19 @@ Planned feature improvements
 * Use pg_basebackup for the data directory, and ALSO rsync for the
   configuration files.
 
-* Use pg_basebackup -X s
+* If no configuration file supplied, search in sensible default locations
-  NOTE: this can be used by including `-X s` in the configuration parameter
+  (currently: current directory and `pg_config --sysconfdir`); if
-  `pg_basebackup_options`
+  possible this should include the location provided by the package,
+  if installed.
+
+* repmgrd: if connection to the upstream node fails on startup, optionally
+  retry for a certain period before giving up; this will cover cases when
+  e.g. primary and standby are both starting up, and the standby comes up
+  before the primary. See github issue #80.
+
+* make old master node ID available for event notification commands
+  (See github issue #80).
+
+* Have pg_basebackup use replication slots, if and when support for
+  this is added; see:
+    http://www.postgresql.org/message-id/555DD2B2.7020000@gmx.net

check_dir.c

@@ -23,14 +23,19 @@
 #include <errno.h>
 #include <stdio.h>
 #include <string.h>
+#include <ftw.h>
 
 /* NB: postgres_fe must be included BEFORE check_dir */
-#include "postgres_fe.h"
+#include <libpq-fe.h>
-#include "check_dir.h"
+#include <postgres_fe.h>
 
+#include "check_dir.h"
 #include "strutil.h"
 #include "log.h"
 
+static bool _create_pg_dir(char *dir, bool force, bool for_witness);
+static int unlink_dir_callback(const char *fpath, const struct stat *sb, int typeflag, struct FTW *ftwbuf);
+
 /*
  * make sure the directory either doesn't exist or is empty
  * we use this function to check the new data directory and
@@ -243,6 +248,19 @@ is_pg_dir(char *dir)
 
 bool
 create_pg_dir(char *dir, bool force)
+{
+	return _create_pg_dir(dir, force, false);
+}
+
+bool
+create_witness_pg_dir(char *dir, bool force)
+{
+	return _create_pg_dir(dir, force, true);
+}
+
+
+static bool
+_create_pg_dir(char *dir, bool force, bool for_witness)
 {
 	bool		pg_dir = false;
 
@@ -279,12 +297,24 @@ create_pg_dir(char *dir, bool force)
 
 			pg_dir = is_pg_dir(dir);
 
-			/*
+
-			 * we use force to reduce the time needed to restore a node which
-			 * turn async after a failover or anything else
-			 */
 			if (pg_dir && force)
 			{
+
+				/*
+				 * The witness server does not store any data other than a copy of the
+				 * repmgr metadata, so in --force mode we can simply overwrite the
+				 * directory.
+				 *
+				 * For non-witness servers, we'll leave the data in place, both to reduce
+				 * the risk of unintentional data loss and to make it possible for the
+				 * data directory to be brought up-to-date with rsync.
+				 */
+				if (for_witness)
+				{
+					log_notice(_("deleting existing data directory \"%s\"\n"), dir);
+					nftw(dir, unlink_dir_callback, 64, FTW_DEPTH | FTW_PHYS);
+				}
 				/* Let it continue */
 				break;
 			}
@@ -306,3 +336,14 @@ create_pg_dir(char *dir, bool force)
 	}
 	return true;
 }
+
+static int
+unlink_dir_callback(const char *fpath, const struct stat *sb, int typeflag, struct FTW *ftwbuf)
+{
+    int rv = remove(fpath);
+
+    if (rv)
+        perror(fpath);
+
+    return rv;
+}

check_dir.h

@@ -26,5 +26,6 @@ bool		create_dir(char *dir);
 bool		set_dir_permissions(char *dir);
 bool		is_pg_dir(char *dir);
 bool		create_pg_dir(char *dir, bool force);
+bool		create_witness_pg_dir(char *dir, bool force);
 
 #endif

config.c

@@ -27,9 +27,11 @@
 static void parse_event_notifications_list(t_configuration_options *options, const char *arg);
 static void tablespace_list_append(t_configuration_options *options, const char *arg);
 
+static char config_file_path[MAXPGPATH];
+static bool config_file_provided = false;
 
 /*
- * parse_config()
+ * load_config()
  *
  * Set default options and overwrite with values from provided configuration
  * file.
@@ -40,30 +42,21 @@ static void tablespace_list_append(t_configuration_options *options, const char
  * reload_config()
  */
 bool
-parse_config(const char *config_file, t_configuration_options *options)
+load_config(const char *config_file, t_configuration_options *options, char *argv0)
 {
-	char	   *s,
+	struct stat config;
-				buff[MAXLINELENGTH];
-	char		config_file_buf[MAXLEN];
-	char		name[MAXLEN];
-	char		value[MAXLEN];
-	bool 		config_file_provided = false;
-	FILE	   *fp;
-
 	/* Sanity checks */
 
 	/*
 	 * If a configuration file was provided, check it exists, otherwise
-	 * emit an error
+	 * emit an error and terminate
 	 */
 	if (config_file[0])
 	{
-		struct stat config;
+		strncpy(config_file_path, config_file, MAXPGPATH);
+		canonicalize_path(config_file_path);
 
-		strncpy(config_file_buf, config_file, MAXLEN);
+		if (stat(config_file_path, &config) != 0)
-		canonicalize_path(config_file_buf);
-
-		if(stat(config_file_buf, &config) != 0)
 		{
 			log_err(_("provided configuration file '%s' not found: %s\n"),
 					config_file,
@@ -76,16 +69,53 @@ parse_config(const char *config_file, t_configuration_options *options)
 	}
 
 	/*
-	 * If no configuration file was provided, set to a default file
+	 * If no configuration file was provided, attempt to find a default file
-	 * which `parse_config()` will attempt to read if it exists
 	 */
-	else
+	if (config_file_provided == false)
 	{
-		strncpy(config_file_buf, DEFAULT_CONFIG_FILE, MAXLEN);
+		char		my_exec_path[MAXPGPATH];
+		char		etc_path[MAXPGPATH];
+
+		/* First check if one is in the default sysconfdir */
+		if (find_my_exec(argv0, my_exec_path) < 0)
+		{
+			fprintf(stderr, _("%s: could not find own program executable\n"), argv0);
+			exit(EXIT_FAILURE);
+		}
+
+		get_etc_path(my_exec_path, etc_path);
+
+		snprintf(config_file_path, MAXPGPATH, "%s/repmgr.conf", etc_path);
+
+		log_debug(_("Looking for configuration file in %s\n"), etc_path);
+
+		if (stat(config_file_path, &config) != 0)
+		{
+			/* Not found - default to ./repmgr.conf */
+			strncpy(config_file_path, DEFAULT_CONFIG_FILE, MAXPGPATH);
+			canonicalize_path(config_file_path);
+			log_debug(_("Looking for configuration file in %s\n"), config_file_path);
+		}
+	}
+
+	return parse_config(options);
 }
 
 
-	fp = fopen(config_file_buf, "r");
+bool
+parse_config(t_configuration_options *options)
+{
+	FILE	   *fp;
+	char	   *s,
+				buff[MAXLINELENGTH];
+	char		name[MAXLEN];
+	char		value[MAXLEN];
+
+	/* For sanity-checking provided conninfo string */
+	PQconninfoOption *conninfo_options;
+	char       *conninfo_errmsg = NULL;
+
+	fp = fopen(config_file_path, "r");
 
 	/*
 	 * Since some commands don't require a config file at all, not having one
@@ -101,7 +131,7 @@ parse_config(const char *config_file, t_configuration_options *options)
 	{
 		if (config_file_provided)
 		{
-			log_err(_("unable to open provided configuration file '%s'; terminating\n"), config_file_buf);
+			log_err(_("unable to open provided configuration file '%s'; terminating\n"), config_file_path);
 			exit(ERR_BAD_CONFIG);
 		}
 
@@ -149,13 +179,17 @@ parse_config(const char *config_file, t_configuration_options *options)
 	{
 		bool known_parameter = true;
 
-		/* Skip blank lines and comments */
-		if (buff[0] == '\n' || buff[0] == '#')
-			continue;
-
 		/* Parse name/value pair from line */
 		parse_line(buff, name, value);
 
+		/* Skip blank lines */
+		if (!strlen(name))
+			continue;
+
+		/* Skip comments */
+		if (name[0] == '#')
+			continue;
+
 		/* Copy into correct entry in parameters struct */
 		if (strcmp(name, "cluster") == 0)
 			strncpy(options->cluster_name, value, MAXLEN);
@@ -262,6 +296,12 @@ parse_config(const char *config_file, t_configuration_options *options)
 		exit(ERR_BAD_CONFIG);
 	}
 
+	if (options->node == 0)
+	{
+		log_err(_("'node' must be an integer greater than zero\n"));
+		exit(ERR_BAD_CONFIG);
+	}
+
 	if (*options->node_name == '\0')
 	{
 		log_err(_("required parameter 'node_name' was not found\n"));
@@ -274,6 +314,19 @@ parse_config(const char *config_file, t_configuration_options *options)
 		exit(ERR_BAD_CONFIG);
 	}
 
+	/* Sanity check the provided conninfo string
+	 *
+	 * NOTE: this verifies the string format and checks for valid options
+	 * but does not sanity check values
+	 */
+	conninfo_options = PQconninfoParse(options->conninfo, &conninfo_errmsg);
+	if (conninfo_options == NULL)
+	{
+		log_err(_("Parameter 'conninfo' is invalid: %s"), conninfo_errmsg);
+		exit(ERR_BAD_CONFIG);
+	}
+	PQconninfoFree(conninfo_options);
+
 	/* The following checks are for valid parameter values */
 	if (options->master_response_timeout <= 0)
 	{
@@ -331,24 +384,50 @@ parse_line(char *buff, char *name, char *value)
 	int			j = 0;
 
 	/*
-	 * first we find the name of the parameter
+	 * Extract parameter name, if present
 	 */
 	for (; i < MAXLEN; ++i)
 	{
-		if (buff[i] != '=')
+
-			name[j++] = buff[i];
+		if (buff[i] == '=')
-		else
 			break;
+
+		switch(buff[i])
+		{
+			/* Ignore whitespace */
+			case ' ':
+			case '\n':
+			case '\r':
+			case '\t':
+				continue;
+			default:
+				name[j++] = buff[i];
+		}
 	}
 	name[j] = '\0';
 
 	/*
-	 * Now the value
+	 * Ignore any whitespace following the '=' sign
+	 */
+	for (; i < MAXLEN; ++i)
+	{
+		if (buff[i+1] == ' ')
+			continue;
+		if (buff[i+1] == '\t')
+			continue;
+
+		break;
+	}
+
+	/*
+	 * Extract parameter value
 	 */
 	j = 0;
 	for (++i; i < MAXLEN; ++i)
 		if (buff[i] == '\'')
 			continue;
+		else if (buff[i] == '#')
+			break;
 		else if (buff[i] != '\n')
 			value[j++] = buff[i];
 		else
@@ -358,7 +437,7 @@ parse_line(char *buff, char *name, char *value)
 }
 
 bool
-reload_config(char *config_file, t_configuration_options * orig_options)
+reload_config(t_configuration_options *orig_options)
 {
 	PGconn	   *conn;
 	t_configuration_options new_options;
@@ -369,7 +448,7 @@ reload_config(char *config_file, t_configuration_options * orig_options)
 	 */
 	log_info(_("reloading configuration file and updating repmgr tables\n"));
 
-	parse_config(config_file, &new_options);
+	parse_config(&new_options);
 	if (new_options.node == -1)
 	{
 		log_warning(_("unable to parse new configuration, retaining current configuration\n"));

config.h

@@ -83,9 +83,10 @@ typedef struct
 #define T_CONFIGURATION_OPTIONS_INITIALIZER { "", -1, NO_UPSTREAM_NODE, "", MANUAL_FAILOVER, -1, "", "", "", "", "", "", "", -1, -1, -1, "", "", "", "", 0, 0, 0, "", { NULL, NULL }, {NULL, NULL} }
 
 
-bool		parse_config(const char *config_file, t_configuration_options *options);
+bool		load_config(const char *config_file, t_configuration_options *options, char *argv0);
+bool		reload_config(t_configuration_options *orig_options);
+bool		parse_config(t_configuration_options *options);
 void		parse_line(char *buff, char *name, char *value);
 char	   *trim(char *s);
-bool		reload_config(char *config_file, t_configuration_options *orig_options);
 
 #endif

dbutils.c

@@ -82,6 +82,72 @@ establish_db_connection_by_params(const char *keywords[], const char *values[],
 }
 
 
+bool
+begin_transaction(PGconn *conn)
+{
+	PGresult   *res;
+
+	res = PQexec(conn, "BEGIN");
+
+	if (PQresultStatus(res) != PGRES_COMMAND_OK)
+	{
+		log_err(_("Unable to begin transaction: %s\n"),
+				PQerrorMessage(conn));
+
+		PQclear(res);
+		return false;
+	}
+
+	PQclear(res);
+
+	return true;
+}
+
+
+bool
+commit_transaction(PGconn *conn)
+{
+	PGresult   *res;
+
+	res = PQexec(conn, "COMMIT");
+
+	if (PQresultStatus(res) != PGRES_COMMAND_OK)
+	{
+		log_err(_("Unable to commit transaction: %s\n"),
+				PQerrorMessage(conn));
+		PQclear(res);
+
+		return false;
+	}
+
+	PQclear(res);
+
+	return true;
+}
+
+
+bool
+rollback_transaction(PGconn *conn)
+{
+	PGresult   *res;
+
+	res = PQexec(conn, "ROLLBACK");
+
+	if (PQresultStatus(res) != PGRES_COMMAND_OK)
+	{
+		log_err(_("Unable to rollback transaction: %s\n"),
+				PQerrorMessage(conn));
+		PQclear(res);
+
+		return false;
+	}
+
+	PQclear(res);
+
+	return true;
+}
+
+
 bool
 check_cluster_schema(PGconn *conn)
 {
@@ -1080,8 +1146,8 @@ delete_node_record(PGconn *conn, int node, char *action)
  *
  * Note this function may be called with `conn` set to NULL in cases where
  * the master node is not available and it's therefore not possible to write
- * an event record. In this case, if `event_notification_command` is set a user-
+ * an event record. In this case, if `event_notification_command` is set, a
- * defined notification to be generated; if not, this function will have
+ * user-defined notification to be generated; if not, this function will have
  * no effect.
  */
 
@@ -1094,7 +1160,7 @@ create_event_record(PGconn *conn, t_configuration_options *options, int node_id,
 	bool		success = true;
 	struct tm	ts;
 
-	/* Only attempt to write a record if a connection handle was provided/
+	/* Only attempt to write a record if a connection handle was provided.
 	   Also check that the repmgr schema has been properly intialised - if
 	   not it means no configuration file was provided, which can happen with
 	   e.g. `repmgr standby clone`, and we won't know which schema to write to.
@@ -1283,3 +1349,56 @@ create_event_record(PGconn *conn, t_configuration_options *options, int node_id,
 
 	return success;
 }
+
+bool
+update_node_record_set_upstream(PGconn *conn, char *cluster_name, int this_node_id, int new_upstream_node_id)
+{
+	PGresult   *res;
+	char		sqlquery[QUERY_STR_LEN];
+
+	log_debug(_("update_node_record_set_upstream(): Updating node %i's upstream node to %i\n"), this_node_id, new_upstream_node_id);
+
+	sqlquery_snprintf(sqlquery,
+					  "  UPDATE %s.repl_nodes "
+					  "     SET upstream_node_id = %i "
+					  "   WHERE cluster = '%s' "
+					  "     AND id = %i ",
+					  get_repmgr_schema_quoted(conn),
+					  new_upstream_node_id,
+					  cluster_name,
+					  this_node_id);
+	res = PQexec(conn, sqlquery);
+
+	if (PQresultStatus(res) != PGRES_COMMAND_OK)
+	{
+		log_err(_("Unable to set new upstream node id: %s\n"),
+				PQerrorMessage(conn));
+		PQclear(res);
+
+		return false;
+	}
+
+	PQclear(res);
+
+	return true;
+}
+
+
+PGresult *
+get_node_record(PGconn *conn, char *cluster, int node_id)
+{
+	char		sqlquery[QUERY_STR_LEN];
+
+	sprintf(sqlquery,
+			"SELECT id, upstream_node_id, conninfo, type, slot_name, active "
+			"  FROM %s.repl_nodes "
+			" WHERE cluster = '%s' "
+			"   AND id = %i",
+			get_repmgr_schema_quoted(conn),
+			cluster,
+			node_id);
+
+	log_debug("get_node_record(): %s\n", sqlquery);
+
+	return PQexec(conn, sqlquery);
+}

dbutils.h

@@ -30,6 +30,9 @@ PGconn *establish_db_connection(const char *conninfo,
 PGconn *establish_db_connection_by_params(const char *keywords[],
 								  const char *values[],
 								  const bool exit_on_error);
+bool		begin_transaction(PGconn *conn);
+bool		commit_transaction(PGconn *conn);
+bool		rollback_transaction(PGconn *conn);
 bool		check_cluster_schema(PGconn *conn);
 int			is_standby(PGconn *conn);
 bool		is_pgup(PGconn *conn, int timeout);
@@ -63,6 +66,7 @@ bool		copy_configuration(PGconn *masterconn, PGconn *witnessconn, char *cluster_
 bool		create_node_record(PGconn *conn, char *action, int node, char *type, int upstream_node, char *cluster_name, char *node_name, char *conninfo, int priority, char *slot_name);
 bool		delete_node_record(PGconn *conn, int node, char *action);
 bool        create_event_record(PGconn *conn, t_configuration_options *options, int node_id, char *event, bool successful, char *details);
-
+bool        update_node_record_set_upstream(PGconn *conn, char *cluster_name, int this_node_id, int new_upstream_node_id);
+PGresult *	get_node_record(PGconn *conn, char *cluster, int node_id);
 
 #endif

debian/repmgr.repmgrd.default

@@ -12,7 +12,7 @@ REPMGRD_ENABLED=no
 #REPMGRD_USER=postgres
 
 # repmgrd binary
-#REPMGR_BIN=/usr/bin/repmgr
+#REPMGRD_BIN=/usr/bin/repmgrd
 
 # pid file
-#REPMGR_PIDFILE=/var/run/repmgrd.pid
+#REPMGRD_PIDFILE=/var/run/repmgrd.pid

errcode.h

@@ -35,5 +35,6 @@
 #define ERR_BAD_SSH 12
 #define ERR_SYS_FAILURE 13
 #define ERR_BAD_BASEBACKUP 14
+#define ERR_INTERNAL 15
 
 #endif   /* _ERRCODE_H_ */

log.c

@@ -144,12 +144,32 @@ logger_init(t_configuration_options * opts, const char *ident, const char *level
 	{
 		FILE	   *fd;
 
-		fd = freopen(opts->logfile, "a", stderr);
+		/* Check if we can write to the specified file before redirecting
+		 * stderr - if freopen() fails, stderr output will vanish into
+		 * the ether and the user won't know what's going on.
+		 */
 
+		fd = fopen(opts->logfile, "a");
 		if (fd == NULL)
 		{
-			fprintf(stderr, "error reopening stderr to '%s': %s",
+			stderr_log_err(_("Unable to open specified logfile '%s' for writing: %s\n"), opts->logfile, strerror(errno));
-					opts->logfile, strerror(errno));
+			stderr_log_err(_("Terminating\n"));
+			exit(ERR_BAD_CONFIG);
+		}
+		fclose(fd);
+
+		stderr_log_notice(_("Redirecting logging output to '%s'\n"), opts->logfile);
+		fd = freopen(opts->logfile, "a", stderr);
+
+		/* It's possible freopen() may still fail due to e.g. a race condition;
+		   as it's not feasible to restore stderr after a failed freopen(),
+		   we'll write to stdout as a last resort.
+		 */
+		if (fd == NULL)
+		{
+			printf(_("Unable to open specified logfile %s for writing: %s\n"), opts->logfile, strerror(errno));
+			printf(_("Terminating\n"));
+			exit(ERR_BAD_CONFIG);
 		}
 	}
 

repmgr.c

@@ -7,18 +7,19 @@
  *
  * Commands implemented are:
  *
- * MASTER REGISTER
+ * [ MASTER | PRIMARY ] REGISTER
  *
  * STANDBY REGISTER
+ * STANDBY UNREGISTER
  * STANDBY CLONE
  * STANDBY FOLLOW
  * STANDBY PROMOTE
  *
+ * WITNESS CREATE
+ *
  * CLUSTER SHOW
  * CLUSTER CLEANUP
  *
- * WITNESS CREATE
- *
  * This program is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation, either version 3 of the License, or
@@ -53,15 +54,23 @@
 
 #define RECOVERY_FILE "recovery.conf"
 
+#ifndef TABLESPACE_MAP
+#define TABLESPACE_MAP "tablespace_map"
+#endif
+
+#define WITNESS_DEFAULT_PORT "5499" /* If this value is ever changed, remember
+									 * to update comments and documentation */
+
 #define NO_ACTION			0		/* Dummy default action */
 #define MASTER_REGISTER		1
 #define STANDBY_REGISTER	2
-#define STANDBY_CLONE	 3
+#define STANDBY_UNREGISTER	3
-#define STANDBY_PROMOTE  4
+#define STANDBY_CLONE		4
-#define STANDBY_FOLLOW	 5
+#define STANDBY_PROMOTE		5
-#define WITNESS_CREATE	 6
+#define STANDBY_FOLLOW		6
-#define CLUSTER_SHOW	 7
+#define WITNESS_CREATE		7
-#define CLUSTER_CLEANUP  8
+#define CLUSTER_SHOW		8
+#define CLUSTER_CLEANUP		9
 
 
 
@@ -69,7 +78,7 @@ static bool create_recovery_file(const char *data_dir);
 static int	test_ssh_connection(char *host, char *remote_user);
 static int  copy_remote_files(char *host, char *remote_user, char *remote_path,
 							  char *local_path, bool is_directory, int server_version_num);
-static int  run_basebackup(void);
+static int  run_basebackup(const char *data_dir);
 static void check_parameters_for_action(const int action);
 static bool create_schema(PGconn *conn);
 static void write_primary_conninfo(char *line);
@@ -83,6 +92,7 @@ static char *make_pg_path(char *file);
 
 static void do_master_register(void);
 static void do_standby_register(void);
+static void do_standby_unregister(void);
 static void do_standby_clone(void);
 static void do_standby_promote(void);
 static void do_standby_follow(void);
@@ -129,7 +139,7 @@ main(int argc, char **argv)
 		{"port", required_argument, NULL, 'p'},
 		{"username", required_argument, NULL, 'U'},
 		{"superuser", required_argument, NULL, 'S'},
-		{"dest-dir", required_argument, NULL, 'D'},
+		{"data-dir", required_argument, NULL, 'D'},
 		{"local-port", required_argument, NULL, 'l'},
 		{"config-file", required_argument, NULL, 'f'},
 		{"remote-user", required_argument, NULL, 'R'},
@@ -145,6 +155,8 @@ main(int argc, char **argv)
 		{"check-upstream-config", no_argument, NULL, 2},
 		{"recovery-min-apply-delay", required_argument, NULL, 3},
 		{"ignore-external-config-files", no_argument, NULL, 4},
+		{"help", no_argument, NULL, '?'},
+		{"version", no_argument, NULL, 'V'},
 		{NULL, 0, NULL, 0}
 	};
 
@@ -158,28 +170,20 @@ main(int argc, char **argv)
 
 	progname = get_progname(argv[0]);
 
-	if (argc > 1)
-	{
-		if (strcmp(argv[1], "--help") == 0 || strcmp(argv[1], "-?") == 0)
-		{
-			help(progname);
-			exit(SUCCESS);
-		}
-		if (strcmp(argv[1], "--version") == 0 || strcmp(argv[1], "-V") == 0)
-		{
-			printf("%s %s (PostgreSQL %s)\n", progname, REPMGR_VERSION, PG_VERSION);
-			exit(SUCCESS);
-		}
-	}
-
 	/* Prevent getopt_long() from printing an error message */
 	opterr = 0;
 
-	while ((c = getopt_long(argc, argv, "d:h:p:U:S:D:l:f:R:w:k:FWIvb:r:c", long_options,
+	while ((c = getopt_long(argc, argv, "?Vd:h:p:U:S:D:l:f:R:w:k:FWIvb:r:c", long_options,
 							&optindex)) != -1)
 	{
 		switch (c)
 		{
+			case '?':
+				help(progname);
+				exit(SUCCESS);
+			case 'V':
+				printf("%s %s (PostgreSQL %s)\n", progname, REPMGR_VERSION, PG_VERSION);
+				exit(SUCCESS);
 			case 'd':
 				strncpy(runtime_options.dbname, optarg, MAXLEN);
 				break;
@@ -299,8 +303,10 @@ main(int argc, char **argv)
 
 	/*
 	 * Now we need to obtain the action, this comes in one of these forms:
-	 * MASTER REGISTER | STANDBY {REGISTER | CLONE [node] | PROMOTE | FOLLOW
+	 *   MASTER REGISTER |
-	 * [node]} | WITNESS CREATE CLUSTER {SHOW | CLEANUP}
+	 *   STANDBY {REGISTER | UNREGISTER | CLONE [node] | PROMOTE | FOLLOW [node]} |
+	 *   WITNESS CREATE |
+	 *   CLUSTER {SHOW | CLEANUP}
 	 *
 	 * the node part is optional, if we receive it then we shouldn't have
 	 * received a -h option
@@ -310,6 +316,8 @@ main(int argc, char **argv)
 		server_mode = argv[optind++];
 		if (strcasecmp(server_mode, "STANDBY") != 0 &&
 			strcasecmp(server_mode, "MASTER") != 0 &&
+			/* allow PRIMARY as synonym for MASTER */
+			strcasecmp(server_mode, "PRIMARY") != 0 &&
 			strcasecmp(server_mode, "WITNESS") != 0 &&
 			strcasecmp(server_mode, "CLUSTER") != 0)
 		{
@@ -324,7 +332,7 @@ main(int argc, char **argv)
 	{
 		server_cmd = argv[optind++];
 		/* check posibilities for all server modes */
-		if (strcasecmp(server_mode, "MASTER") == 0)
+		if (strcasecmp(server_mode, "MASTER") == 0 || strcasecmp(server_mode, "PRIMARY") == 0 )
 		{
 			if (strcasecmp(server_cmd, "REGISTER") == 0)
 				action = MASTER_REGISTER;
@@ -333,6 +341,8 @@ main(int argc, char **argv)
 		{
 			if (strcasecmp(server_cmd, "REGISTER") == 0)
 				action = STANDBY_REGISTER;
+			if (strcasecmp(server_cmd, "UNREGISTER") == 0)
+				action = STANDBY_UNREGISTER;
 			else if (strcasecmp(server_cmd, "CLONE") == 0)
 				action = STANDBY_CLONE;
 			else if (strcasecmp(server_cmd, "PROMOTE") == 0)
@@ -425,7 +435,6 @@ main(int argc, char **argv)
 
 	if (runtime_options.verbose && runtime_options.config_file[0])
 	{
-
 		log_notice(_("opening configuration file: %s\n"),
 				   runtime_options.config_file);
 	}
@@ -435,7 +444,7 @@ main(int argc, char **argv)
 	 * however if available we'll parse it anyway for options like 'log_level',
 	 * 'use_replication_slots' etc.
 	 */
-	config_file_parsed = parse_config(runtime_options.config_file, &options);
+	config_file_parsed = load_config(runtime_options.config_file, &options, argv[0]);
 
 	/*
 	 * Initialise pg_bindir - command line parameter will override
@@ -542,6 +551,9 @@ main(int argc, char **argv)
 		case STANDBY_REGISTER:
 			do_standby_register();
 			break;
+		case STANDBY_UNREGISTER:
+			do_standby_unregister();
+			break;
 		case STANDBY_CLONE:
 			do_standby_clone();
 			break;
@@ -691,6 +703,7 @@ static void
 do_master_register(void)
 {
 	PGconn	   *conn;
+	PGconn	   *master_conn;
 
 	bool		schema_exists = false;
 	int			ret;
@@ -716,56 +729,74 @@ do_master_register(void)
 		exit(ERR_BAD_CONFIG);
 	}
 
-	/* Check if there is a schema for this cluster */
+	/* Create schema and associated database objects, if it does not exist */
 	schema_exists = check_cluster_schema(conn);
 
-	/* If schema exists and force option not selected, raise an error */
-	if(schema_exists && !runtime_options.force)
-	{
-		log_notice(_("schema '%s' already exists.\n"), get_repmgr_schema());
-		PQfinish(conn);
-		exit(ERR_BAD_CONFIG);
-	}
-
 	if (!schema_exists)
 	{
 		log_info(_("master register: creating database objects inside the %s schema\n"),
 				 get_repmgr_schema());
 
-		/* ok, create the schema */
+		begin_transaction(conn);
+
 		if (!create_schema(conn))
-			return;
-	}
-	else
 		{
-		PGconn	   *master_conn;
+			log_err(_("Unable to create repmgr schema - see preceding error message(s); aborting\n"));
+			rollback_transaction(conn);
+			PQfinish(conn);
+			exit(ERR_BAD_CONFIG);
+		}
+
+		commit_transaction(conn);
+	}
 
 	/* Ensure there isn't any other master already registered */
 	master_conn = get_master_connection(conn,
 										options.cluster_name, NULL, NULL);
-		if (master_conn != NULL)
+
+	if (master_conn != NULL && !runtime_options.force)
 	{
 		PQfinish(master_conn);
-			log_warning(_("there is a master already in cluster %s\n"),
+		log_err(_("there is a master already in cluster %s\n"),
 					options.cluster_name);
 		exit(ERR_BAD_CONFIG);
 	}
 
+	PQfinish(master_conn);
+
+	/* XXX we should check if a node with a different ID is registered as
+	   master, otherwise it would be possible to insert a duplicate record
+	   with --force, which would result in an unwelcome "multi-master" situation
+	*/
+
+	/* Delete any existing record for this node if --force set */
 	if (runtime_options.force)
 	{
-			bool node_record_deleted = delete_node_record(conn,
+		PGresult *res;
+		bool node_record_deleted;
+
+		begin_transaction(conn);
+
+		res = get_node_record(conn, options.cluster_name, options.node);
+		if (PQntuples(res))
+		{
+			log_notice(_("deleting existing master record with id %i\n"), options.node);
+
+			node_record_deleted = delete_node_record(conn,
 													 options.node,
 													 "master register");
-
 			if (node_record_deleted == false)
 			{
-				PQfinish(master_conn);
+				rollback_transaction(conn);
 				PQfinish(conn);
 				exit(ERR_BAD_CONFIG);
 			}
 		}
+
+		commit_transaction(conn);
 	}
 
+
 	/* Now register the master */
 	record_created = create_node_record(conn,
 										"master register",
@@ -908,6 +939,85 @@ do_standby_register(void)
 }
 
 
+static void
+do_standby_unregister(void)
+{
+	PGconn	   *conn;
+	PGconn	   *master_conn;
+	int			ret;
+
+	bool		node_record_deleted;
+
+	log_info(_("connecting to standby database\n"));
+	conn = establish_db_connection(options.conninfo, true);
+
+	/* Check we are a standby */
+	ret = is_standby(conn);
+	if (ret == 0 || ret == -1)
+	{
+		log_err(_(ret == 0 ? "this node should be a standby (%s)\n" :
+				"connection to node (%s) lost\n"), options.conninfo);
+
+		PQfinish(conn);
+		exit(ERR_BAD_CONFIG);
+	}
+
+	/* Check if there is a schema for this cluster */
+	if (check_cluster_schema(conn) == false)
+	{
+		/* schema doesn't exist */
+		log_err(_("schema '%s' doesn't exist.\n"), get_repmgr_schema());
+		PQfinish(conn);
+		exit(ERR_BAD_CONFIG);
+	}
+
+	/* check if there is a master in this cluster */
+	log_info(_("connecting to master database\n"));
+	master_conn = get_master_connection(conn, options.cluster_name,
+										NULL, NULL);
+	if (!master_conn)
+	{
+		log_err(_("a master must be defined before unregistering a slave\n"));
+		exit(ERR_BAD_CONFIG);
+	}
+
+	/*
+	 * Verify that standby and master are supported and compatible server
+	 * versions
+	 */
+	check_master_standby_version_match(conn, master_conn);
+
+	/* Now unregister the standby */
+	log_info(_("unregistering the standby\n"));
+	node_record_deleted = delete_node_record(master_conn,
+										     options.node,
+											 "standby unregister");
+
+	if (node_record_deleted == false)
+	{
+		PQfinish(master_conn);
+		PQfinish(conn);
+		exit(ERR_BAD_CONFIG);
+	}
+
+	/* Log the event */
+	create_event_record(master_conn,
+						&options,
+						options.node,
+						"standby_unregister",
+						true,
+						NULL);
+
+	PQfinish(master_conn);
+	PQfinish(conn);
+
+	log_info(_("standby unregistration complete\n"));
+	log_notice(_("standby node correctly unregistered for cluster %s with id %d (conninfo: %s)\n"),
+			   options.cluster_name, options.node, options.conninfo);
+	return;
+}
+
+
 static void
 do_standby_clone(void)
 {
@@ -1009,14 +1119,14 @@ do_standby_clone(void)
 	 *
 	 * -T/--tablespace-mapping is not available as a pg_basebackup option for
 	 * PostgreSQL 9.3 - we can only handle that with rsync, so if `--rsync-only`
-	 # not set, fail with an error
+	 * not set, fail with an error
 	 */
 
 	if (options.tablespace_mapping.head != NULL)
 	{
 		TablespaceListCell *cell;
 
-		if(get_server_version(upstream_conn, NULL) < 90400)
+		if (get_server_version(upstream_conn, NULL) < 90400 && !runtime_options.rsync_only)
 		{
 			log_err(_("in PostgreSQL 9.3, tablespace mapping can only be used in conjunction with --rsync-only\n"));
 			PQfinish(upstream_conn);
@@ -1184,6 +1294,15 @@ do_standby_clone(void)
 
 	if (runtime_options.rsync_only)
 	{
+		PQExpBufferData tablespace_map;
+		bool		tablespace_map_rewrite = false;
+
+		/* For 9.5 and greater, create our own tablespace_map file */
+		if (server_version_num >= 90500)
+		{
+			initPQExpBuffer(&tablespace_map);
+		}
+
 		/*
 		 * From pg 9.1 default is to wait for a sync standby to ack, avoid that by
 		 * turning off sync rep for this session
@@ -1258,6 +1377,7 @@ do_standby_clone(void)
 			initPQExpBuffer(&tblspc_dir_dst);
 			initPQExpBuffer(&tblspc_oid);
 
+
 			appendPQExpBuffer(&tblspc_oid, "%s", PQgetvalue(res, i, 0));
 			appendPQExpBuffer(&tblspc_dir_src, "%s", PQgetvalue(res, i, 1));
 
@@ -1292,8 +1412,21 @@ do_standby_clone(void)
 								  tblspc_dir_src.data, tblspc_dir_dst.data,
 								  true, server_version_num);
 
-			/* Update symlink in pg_tblspc */
+
+			/* Update symlinks in pg_tblspc */
 			if (mapping_found == true)
+			{
+				/* 9.5 and later - create a tablespace_map file */
+				if (server_version_num >= 90500)
+				{
+					tablespace_map_rewrite = true;
+					appendPQExpBuffer(&tablespace_map,
+									  "%s %s\n",
+									  tblspc_oid.data,
+									  tblspc_dir_dst.data);
+				}
+				/* Pre-9.5, we have to manipulate the symlinks in pg_tblspc/ ourselves */
+				else
 				{
 					PQExpBufferData tblspc_symlink;
 
@@ -1315,13 +1448,49 @@ do_standby_clone(void)
 
 
 				}
+
+			}
+		}
+
+
+		if(server_version_num >= 90500 && tablespace_map_rewrite == true)
+		{
+			PQExpBufferData tablespace_map_filename;
+			FILE	   *tablespace_map_file;
+			initPQExpBuffer(&tablespace_map_filename);
+			appendPQExpBuffer(&tablespace_map_filename, "%s/%s",
+							  local_data_directory,
+							  TABLESPACE_MAP);
+
+			/* Unlink any existing file (it should be there, but we don't care if it isn't) */
+			if (unlink(tablespace_map_filename.data) < 0 && errno != ENOENT)
+			{
+				log_err(_("unable to remove tablespace_map file %s\n"), tablespace_map_filename.data);
+				exit(ERR_BAD_CONFIG);
+			}
+
+			tablespace_map_file = fopen(tablespace_map_filename.data, "w");
+			if (tablespace_map_file == NULL)
+			{
+				log_err(_("unable to create tablespace_map file '%s'\n"), tablespace_map_filename.data);
+				exit(ERR_BAD_CONFIG);
+			}
+
+			if (fputs(tablespace_map.data, tablespace_map_file) == EOF)
+			{
+				log_err(_("unable to write to tablespace_map file '%s'\n"), tablespace_map_filename.data);
+				fclose(tablespace_map_file);
+				exit(ERR_BAD_CONFIG);
+			}
+
+			fclose(tablespace_map_file);
 		}
 
 		PQclear(res);
 	}
 	else
 	{
-		r = run_basebackup();
+		r = run_basebackup(local_data_directory);
 		if (r != 0)
 		{
 			log_warning(_("standby clone: base backup failed\n"));
@@ -1449,6 +1618,24 @@ stop_backup:
 		exit(retval);
 	}
 
+	/*
+	 * Remove existing WAL from the target directory, since
+	 * rsync's --exclude option doesn't do it.
+	 */
+	if (runtime_options.force)
+	{
+		char	script[MAXLEN];
+		maxlen_snprintf(script, "rm -rf %s/pg_xlog/*",
+						local_data_directory);
+		r = system(script);
+		if (r != 0)
+		{
+			log_err(_("unable to empty local WAL directory %s/pg_xlog/\n"),
+					local_data_directory);
+			exit(ERR_BAD_RSYNC);
+		}
+	}
+
 	/* Finally, write the recovery.conf file */
 	create_recovery_file(local_data_directory);
 
@@ -1678,6 +1865,7 @@ do_standby_follow(void)
 	char		script[MAXLEN];
 	char		master_conninfo[MAXLEN];
 	PGconn	   *master_conn;
+	int			master_id;
 
 	int			r,
 				retval;
@@ -1716,7 +1904,7 @@ do_standby_follow(void)
 		}
 
 		master_conn = get_master_connection(conn,
-				options.cluster_name, NULL, (char *) &master_conninfo);
+				options.cluster_name, &master_id, (char *) &master_conninfo);
 	}
 	while (master_conn == NULL && runtime_options.wait_for_master);
 
@@ -1753,7 +1941,6 @@ do_standby_follow(void)
 	strncpy(runtime_options.host, PQhost(master_conn), MAXLEN);
 	strncpy(runtime_options.masterport, PQport(master_conn), MAXLEN);
 	strncpy(runtime_options.username, PQuser(master_conn), MAXLEN);
-	PQfinish(master_conn);
 
 	log_info(_("changing standby's master\n"));
 
@@ -1785,6 +1972,16 @@ do_standby_follow(void)
 		exit(ERR_NO_RESTART);
 	}
 
+	if (update_node_record_set_upstream(master_conn, options.cluster_name,
+									   options.node, master_id) == false)
+	{
+		log_err(_("unable to update upstream node"));
+		PQfinish(master_conn);
+
+		exit(ERR_BAD_CONFIG);
+	}
+	PQfinish(master_conn);
+
 	return;
 }
 
@@ -1808,6 +2005,9 @@ do_witness_create(void)
 	bool        success;
 	bool		record_created;
 
+	PQconninfoOption *conninfo_options;
+	PQconninfoOption *conninfo_option;
+
 	/* Connection parameters for master only */
 	keywords[0] = "host";
 	values[0] = runtime_options.host;
@@ -1818,7 +2018,7 @@ do_witness_create(void)
 	masterconn = establish_db_connection_by_params(keywords, values, true);
 	if (!masterconn)
 	{
-		/* No event logging possible as we can't connect to the master */
+		/* No event logging possible here as we can't connect to the master */
 		log_err(_("unable to connect to master\n"));
 		exit(ERR_DB_CON);
 	}
@@ -1873,7 +2073,7 @@ do_witness_create(void)
 	}
 
 	/* Check this directory could be used as a PGDATA dir */
-	if (!create_pg_dir(runtime_options.dest_dir, runtime_options.force))
+	if (!create_witness_pg_dir(runtime_options.dest_dir, runtime_options.force))
 	{
 		PQExpBufferData errmsg;
 		initPQExpBuffer(&errmsg);
@@ -1949,6 +2149,27 @@ do_witness_create(void)
 	xsnprintf(buf, sizeof(buf), "\n#Configuration added by %s\n", progname);
 	fputs(buf, pg_conf);
 
+
+	/* Attempt to extract a port number from the provided conninfo string
+	 * This will override any value provided with '-l/--local-port', as it's
+	 * what we'll later try and connect to anyway. '-l/--local-port' should
+	 * be deprecated.
+	 */
+	conninfo_options = PQconninfoParse(options.conninfo, NULL);
+
+	for (conninfo_option = conninfo_options; conninfo_option->keyword != NULL; conninfo_option++)
+	{
+		if (strcmp(conninfo_option->keyword, "port") == 0)
+		{
+			if (conninfo_option->val != NULL && conninfo_option->val[0] != '\0')
+			{
+				strncpy(runtime_options.localport, conninfo_option->val, MAXLEN);
+				break;
+			}
+		}
+	}
+	PQconninfoFree(conninfo_options);
+
 	/*
 	 * If not specified by the user, the default port for the witness server
 	 * is 5499; this is intended to support running the witness server as
@@ -1956,7 +2177,7 @@ do_witness_create(void)
 	 * dedicated server.
 	 */
 	if (!runtime_options.localport[0])
-		strncpy(runtime_options.localport, "5499", MAXLEN);
+		strncpy(runtime_options.localport, WITNESS_DEFAULT_PORT, MAXLEN);
 
 	xsnprintf(buf, sizeof(buf), "port = %s\n", runtime_options.localport);
 	fputs(buf, pg_conf);
@@ -2105,6 +2326,19 @@ do_witness_create(void)
 
 	/* register ourselves in the master */
 
+	if (runtime_options.force)
+	{
+		bool node_record_deleted = delete_node_record(masterconn,
+													  options.node,
+													  "witness create");
+
+		if (node_record_deleted == false)
+		{
+			PQfinish(masterconn);
+			exit(ERR_BAD_CONFIG);
+		}
+	}
+
 	record_created = create_node_record(masterconn,
 										"witness create",
 										options.node,
@@ -2134,8 +2368,12 @@ do_witness_create(void)
 
 	log_info(_("starting copy of configuration from master...\n"));
 
+	begin_transaction(witnessconn);
+
+
 	if (!create_schema(witnessconn))
 	{
+		rollback_transaction(witnessconn);
 		create_event_record(masterconn,
 							&options,
 							options.node,
@@ -2147,6 +2385,8 @@ do_witness_create(void)
 		exit(ERR_BAD_CONFIG);
 	}
 
+	commit_transaction(witnessconn);
+
 	/* copy configuration from master, only repl_nodes is needed */
 	if (!copy_configuration(masterconn, witnessconn, options.cluster_name))
 	{
@@ -2199,16 +2439,18 @@ do_witness_create(void)
 static void
 help(const char *progname)
 {
-	printf(_("\n%s: Replicator manager \n"), progname);
+	printf(_("%s: replication management tool for PostgreSQL\n"), progname);
+	printf(_("\n"));
 	printf(_("Usage:\n"));
 	printf(_("  %s [OPTIONS] master  {register}\n"), progname);
-	printf(_(" %s [OPTIONS] standby {register|clone|promote|follow}\n"),
+	printf(_("  %s [OPTIONS] standby {register|unregister|clone|promote|follow}\n"),
 		   progname);
 	printf(_("  %s [OPTIONS] cluster {show|cleanup}\n"), progname);
-	printf(_("\nGeneral options:\n"));
+	printf(_("\n"));
-	printf(_("  --help                              show this help, then exit\n"));
+	printf(_("General options:\n"));
-	printf(_("  --version                           output version information, then exit\n"));
+	printf(_("  -?, --help                          show this help, then exit\n"));
-	printf(_("  --verbose                           output verbose activity information\n"));
+	printf(_("  -V, --version                       output version information, then exit\n"));
+	printf(_("  -v, --verbose                       output verbose activity information\n"));
 	printf(_("\nConnection options:\n"));
 	printf(_("  -d, --dbname=DBNAME                 database to connect to\n"));
 	printf(_("  -h, --host=HOSTNAME                 database server host or socket directory\n"));
@@ -2218,11 +2460,12 @@ help(const char *progname)
 	printf(_("  -b, --pg_bindir=PATH                path to PostgreSQL binaries (optional)\n"));
 	printf(_("  -D, --data-dir=DIR                  local directory where the files will be\n" \
 			 "                                      copied to\n"));
-	printf(_("  -l, --local-port=PORT               standby or witness server local port\n"));
 	printf(_("  -f, --config-file=PATH              path to the configuration file\n"));
 	printf(_("  -R, --remote-user=USERNAME          database server username for rsync\n"));
 	printf(_("  -S, --superuser=USERNAME            superuser username for witness database\n" \
 			 "                                      (default: postgres)\n"));
+/* remove this line in the next significant release */
+	printf(_("  -l, --local-port=PORT               (DEPRECATED) witness server local port (default: %s)\n"), WITNESS_DEFAULT_PORT);
 	printf(_("  -w, --wal-keep-segments=VALUE       minimum value for the GUC\n" \
 			 "                                      wal_keep_segments (default: %s)\n"), DEFAULT_WAL_KEEP_SEGMENTS);
 	printf(_("  -k, --keep-history=VALUE            keeps indicated number of days of history\n"));
@@ -2241,8 +2484,10 @@ help(const char *progname)
 	printf(_(" master register         - registers the master in a cluster\n"));
 	printf(_(" standby clone [node]    - creates a new standby\n"));
 	printf(_(" standby register        - registers a standby in a cluster\n"));
+	printf(_(" standby unregister      - unregisters a standby in a cluster\n"));
 	printf(_(" standby promote         - promotes a specific standby to master\n"));
 	printf(_(" standby follow          - makes standby follow a new master\n"));
+	printf(_(" witness create          - creates a new witness server\n"));
 	printf(_(" cluster show            - displays information about cluster nodes\n"));
 	printf(_(" cluster cleanup         - prunes or truncates monitoring history\n" \
 			 "                           (monitoring history creation requires repmgrd\n" \
@@ -2471,17 +2716,19 @@ copy_remote_files(char *host, char *remote_user, char *remote_path,
 
 
 static int
-run_basebackup()
+run_basebackup(const char *data_dir)
 {
 	char				script[MAXLEN];
 	int					r = 0;
 	PQExpBufferData 	params;
 	TablespaceListCell *cell;
 
-	/* Creare pg_basebackup command line options */
+	/* Create pg_basebackup command line options */
 
 	initPQExpBuffer(&params);
 
+	appendPQExpBuffer(&params, " -D %s", data_dir);
+
 	if (strlen(runtime_options.host))
 	{
 		appendPQExpBuffer(&params, " -h %s", runtime_options.host);
@@ -2497,11 +2744,6 @@ run_basebackup()
 		appendPQExpBuffer(&params, " -U %s", runtime_options.username);
 	}
 
-	if(strlen(runtime_options.dest_dir))
-	{
-		appendPQExpBuffer(&params, " -D %s", runtime_options.dest_dir);
-	}
-
 	if (runtime_options.fast_checkpoint) {
 		appendPQExpBuffer(&params, " -c fast");
 	}
@@ -2577,6 +2819,23 @@ check_parameters_for_action(const int action)
 				error_list_append(_("destination directory not required when executing STANDBY REGISTER"));
 			}
 			break;
+		case STANDBY_UNREGISTER:
+
+			/*
+			 * To unregister a standby we only need the repmgr.conf we don't
+			 * need connection parameters to the master because we can detect
+			 * the master in repl_nodes
+			 */
+			if (runtime_options.host[0] || runtime_options.masterport[0] ||
+				runtime_options.username[0] || runtime_options.dbname[0])
+			{
+				error_list_append(_("master connection parameters not required when executing STANDBY UNREGISTER"));
+			}
+			if (runtime_options.dest_dir[0])
+			{
+				error_list_append(_("destination directory not required when executing STANDBY UNREGISTER"));
+			}
+			break;
 		case STANDBY_PROMOTE:
 
 			/*
@@ -2676,6 +2935,7 @@ check_parameters_for_action(const int action)
 }
 
 
+/* The caller should wrap this function in a transaction */
 static bool
 create_schema(PGconn *conn)
 {
@@ -2690,8 +2950,11 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create the schema %s: %s\n"),
 				get_repmgr_schema(), PQerrorMessage(conn));
-		PQfinish(conn);
+
-		exit(ERR_BAD_CONFIG);
+		if (res != NULL)
+			PQclear(res);
+
+		return false;
 	}
 	PQclear(res);
 
@@ -2714,6 +2977,10 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create the function repmgr_update_last_updated: %s\n"),
 				PQerrorMessage(conn));
+
+		if (res != NULL)
+			PQclear(res);
+
 		return false;
 	}
 	PQclear(res);
@@ -2731,6 +2998,10 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create the function repmgr_get_last_updated: %s\n"),
 				PQerrorMessage(conn));
+
+		if (res != NULL)
+			PQclear(res);
+
 		return false;
 	}
 	PQclear(res);
@@ -2759,8 +3030,11 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create table '%s.repl_nodes': %s\n"),
 				get_repmgr_schema_quoted(conn), PQerrorMessage(conn));
-		PQfinish(conn);
+
-		exit(ERR_BAD_CONFIG);
+		if (res != NULL)
+			PQclear(res);
+
+		return false;
 	}
 	PQclear(res);
 
@@ -2783,8 +3057,11 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create table '%s.repl_monitor': %s\n"),
 				get_repmgr_schema_quoted(conn), PQerrorMessage(conn));
-		PQfinish(conn);
+
-		exit(ERR_BAD_CONFIG);
+		if (res != NULL)
+			PQclear(res);
+
+		return false;
 	}
 	PQclear(res);
 
@@ -2806,8 +3083,11 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create table '%s.repl_events': %s\n"),
 				get_repmgr_schema_quoted(conn), PQerrorMessage(conn));
-		PQfinish(conn);
+
-		exit(ERR_BAD_CONFIG);
+		if (res != NULL)
+			PQclear(res);
+
+		return false;
 	}
 	PQclear(res);
 
@@ -2840,8 +3120,11 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create view %s.repl_status: %s\n"),
 				get_repmgr_schema_quoted(conn), PQerrorMessage(conn));
-		PQfinish(conn);
+
-		exit(ERR_BAD_CONFIG);
+		if (res != NULL)
+			PQclear(res);
+
+		return false;
 	}
 	PQclear(res);
 
@@ -2857,8 +3140,11 @@ create_schema(PGconn *conn)
 	{
 		log_err(_("unable to create index 'idx_repl_status_sort' on '%s.repl_monitor': %s\n"),
 				get_repmgr_schema_quoted(conn), PQerrorMessage(conn));
-		PQfinish(conn);
+
-		exit(ERR_BAD_CONFIG);
+		if (res != NULL)
+			PQclear(res);
+
+		return false;
 	}
 	PQclear(res);
 
@@ -2878,6 +3164,8 @@ create_schema(PGconn *conn)
 	{
 		fprintf(stderr, "Cannot create the function repmgr_update_standby_location: %s\n",
 				PQerrorMessage(conn));
+
+		if (res != NULL)
 			PQclear(res);
 
 		return false;
@@ -2896,7 +3184,10 @@ create_schema(PGconn *conn)
 	{
 		fprintf(stderr, "Cannot create the function repmgr_get_last_standby_location: %s\n",
 				PQerrorMessage(conn));
+
+		if (res != NULL)
 			PQclear(res);
+
 		return false;
 	}
 	PQclear(res);
@@ -2945,7 +3236,7 @@ write_primary_conninfo(char *line)
 	}
 
 	maxlen_snprintf(conn_buf, "port=%s%s%s%s%s",
-	   (runtime_options.masterport[0]) ? runtime_options.masterport : "5432",
+	   (runtime_options.masterport[0]) ? runtime_options.masterport : DEF_PGPORT_STR,
 					host_buf, user_buf, password_buf,
 					appname_buf);
 
@@ -3259,18 +3550,7 @@ update_node_record_set_master(PGconn *conn, int this_node_id)
 
 	log_debug(_("Setting %i as master and marking existing master as failed\n"), this_node_id);
 
-	res = PQexec(conn, "BEGIN");
+	begin_transaction(conn);
-
-	if (PQresultStatus(res) != PGRES_COMMAND_OK)
-	{
-		log_err(_("Unable to begin transaction: %s\n"),
-				PQerrorMessage(conn));
-
-		PQclear(res);
-		return false;
-	}
-
-	PQclear(res);
 
 	sqlquery_snprintf(sqlquery,
 					  "  UPDATE %s.repl_nodes "
@@ -3289,7 +3569,7 @@ update_node_record_set_master(PGconn *conn, int this_node_id)
 				PQerrorMessage(conn));
 		PQclear(res);
 
-		PQexec(conn, "ROLLBACK");
+		rollback_transaction(conn);
 		return false;
 	}
 
@@ -3320,20 +3600,7 @@ update_node_record_set_master(PGconn *conn, int this_node_id)
 
 	PQclear(res);
 
-	res = PQexec(conn, "COMMIT");
+	return commit_transaction(conn);
-
-	if (PQresultStatus(res) != PGRES_COMMAND_OK)
-	{
-		log_err(_("Unable to set commit transaction: %s\n"),
-				PQerrorMessage(conn));
-		PQclear(res);
-
-		return false;
-	}
-
-	PQclear(res);
-
-	return true;
 }
 
 
@@ -3344,7 +3611,7 @@ do_check_upstream_config(void)
 	bool		config_ok;
 	int			server_version_num;
 
-	parse_config(runtime_options.config_file, &options);
+	parse_config(&options);
 
 	/* Connection parameters for upstream server only */
 	keywords[0] = "host";

repmgr.h

@@ -20,11 +20,9 @@
 #ifndef _REPMGR_H_
 #define _REPMGR_H_
 
-#include "postgres_fe.h"
+#include <libpq-fe.h>
-#include "libpq-fe.h"
+#include <postgres_fe.h>
-
+#include <getopt_long.h>
-
-#include "getopt_long.h"
 
 #include "strutil.h"
 #include "dbutils.h"

repmgrd.c

@@ -88,11 +88,9 @@ static void check_node_configuration(void);
 
 static void standby_monitor(void);
 static void witness_monitor(void);
-static bool check_connection(PGconn *conn, const char *type);
+static bool check_connection(PGconn **conn, const char *type, const char *conninfo);
 static bool set_local_node_failed(void);
 
-static bool update_node_record_set_upstream(PGconn *conn, int this_node_id, int new_upstream_node_id);
-
 static void update_shared_memory(char *last_wal_standby_applied);
 static void update_registration(void);
 static void do_master_failover(void);
@@ -147,6 +145,8 @@ main(int argc, char **argv)
 		{"monitoring-history", no_argument, NULL, 'm'},
 		{"daemonize", no_argument, NULL, 'd'},
 		{"pid-file", required_argument, NULL, 'p'},
+		{"help", no_argument, NULL, '?'},
+		{"version", no_argument, NULL, 'V'},
 		{NULL, 0, NULL, 0}
 	};
 
@@ -160,21 +160,7 @@ main(int argc, char **argv)
 	int			server_version_num = 0;
 	progname = get_progname(argv[0]);
 
-	if (argc > 1)
+	while ((c = getopt_long(argc, argv, "?Vf:v:mdp:", long_options, &optindex)) != -1)
-	{
-		if (strcmp(argv[1], "--help") == 0 || strcmp(argv[1], "-?") == 0)
-		{
-			help(progname);
-			exit(SUCCESS);
-		}
-		if (strcmp(argv[1], "--version") == 0 || strcmp(argv[1], "-V") == 0)
-		{
-			printf("%s %s (PostgreSQL %s)\n", progname, REPMGR_VERSION, PG_VERSION);
-			exit(SUCCESS);
-		}
-	}
-
-	while ((c = getopt_long(argc, argv, "f:v:mdp:", long_options, &optindex)) != -1)
 	{
 		switch (c)
 		{
@@ -193,6 +179,12 @@ main(int argc, char **argv)
 			case 'p':
 				pid_file = optarg;
 				break;
+			case '?':
+				help(progname);
+				exit(SUCCESS);
+			case 'V':
+				printf("%s %s (PostgreSQL %s)\n", progname, REPMGR_VERSION, PG_VERSION);
+				exit(SUCCESS);
 			default:
 				usage();
 				exit(ERR_BAD_CONFIG);
@@ -208,7 +200,7 @@ main(int argc, char **argv)
 	 * which case we'll need to refactor parse_config() not to abort,
 	 * and return the error message.
 	 */
-	parse_config(config_file, &local_options);
+	load_config(config_file, &local_options, argv[0]);
 
 	if (daemonize)
 	{
@@ -319,7 +311,7 @@ main(int argc, char **argv)
 				check_cluster_configuration(my_local_conn);
 				check_node_configuration();
 
-				if (reload_config(config_file, &local_options))
+				if (reload_config(&local_options))
 				{
 					PQfinish(my_local_conn);
 					my_local_conn = establish_db_connection(local_options.conninfo, true);
@@ -353,7 +345,7 @@ main(int argc, char **argv)
 				 */
 				do
 				{
-					if (check_connection(master_conn, "master"))
+					if (check_connection(&master_conn, "master", NULL))
 					{
 						sleep(local_options.monitor_interval_secs);
 					}
@@ -368,10 +360,10 @@ main(int argc, char **argv)
 					if (got_SIGHUP)
 					{
 						/*
-						 * if we can reload, then could need to change
+						 * if we can reload the configuration file, then could need to change
 						 * my_local_conn
 						 */
-						if (reload_config(config_file, &local_options))
+						if (reload_config(&local_options))
 						{
 							PQfinish(my_local_conn);
 							my_local_conn = establish_db_connection(local_options.conninfo, true);
@@ -432,7 +424,7 @@ main(int argc, char **argv)
 				check_cluster_configuration(my_local_conn);
 				check_node_configuration();
 
-				if (reload_config(config_file, &local_options))
+				if (reload_config(&local_options))
 				{
 					PQfinish(my_local_conn);
 					my_local_conn = establish_db_connection(local_options.conninfo, true);
@@ -483,7 +475,7 @@ main(int argc, char **argv)
 						 * if we can reload, then could need to change
 						 * my_local_conn
 						 */
-						if (reload_config(config_file, &local_options))
+						if (reload_config(&local_options))
 						{
 							PQfinish(my_local_conn);
 							my_local_conn = establish_db_connection(local_options.conninfo, true);
@@ -536,7 +528,7 @@ witness_monitor(void)
 	 * of a missing master and promotion of a standby by that standby's
 	 * repmgrd, so we'll loop for a while before giving up.
 	 */
-	connection_ok = check_connection(master_conn, "master");
+	connection_ok = check_connection(&master_conn, "master", NULL);
 
 	if (connection_ok == false)
 	{
@@ -693,6 +685,7 @@ standby_monitor(void)
 	bool		did_retry = false;
 
 	PGconn	   *upstream_conn;
+	char		upstream_conninfo[MAXCONNINFO];
 	int			upstream_node_id;
 	t_node_info upstream_node;
 
@@ -704,7 +697,7 @@ standby_monitor(void)
 	 * no point in doing much else anyway
 	 */
 
-	if (!check_connection(my_local_conn, "standby"))
+	if (!check_connection(&my_local_conn, "standby", NULL))
 	{
 		PQExpBufferData errmsg;
 
@@ -730,7 +723,7 @@ standby_monitor(void)
 	upstream_conn = get_upstream_connection(my_local_conn,
 											local_options.cluster_name,
 											local_options.node,
-											&upstream_node_id, NULL);
+											&upstream_node_id, upstream_conninfo);
 
 	type = upstream_node_id == master_options.node
 		? "master"
@@ -742,12 +735,12 @@ standby_monitor(void)
 	 * we cannot reconnect, try to get a new upstream node.
 	 */
 
-	check_connection(upstream_conn, type);	/* this takes up to
+	check_connection(&upstream_conn, type, upstream_conninfo);
-											 * local_options.reconnect_attempts
+	/*
+	 * This takes up to local_options.reconnect_attempts *
 	 * local_options.reconnect_intvl seconds
 	 */
 
-
 	if (PQstatus(upstream_conn) != CONNECTION_OK)
 	{
 		PQfinish(upstream_conn);
@@ -879,7 +872,7 @@ standby_monitor(void)
 				log_err(_("standby node has disappeared, trying to reconnect...\n"));
 				did_retry = true;
 
-				if (!check_connection(my_local_conn, "standby"))
+				if (!check_connection(&my_local_conn, "standby", NULL))
 				{
 					set_local_node_failed();
 					terminate(0);
@@ -944,8 +937,9 @@ standby_monitor(void)
 		master_conn = get_master_connection(my_local_conn,
 											 local_options.cluster_name,
 											 &master_options.node, NULL);
-
 	}
+	if (PQstatus(master_conn) != CONNECTION_OK)
+		PQreset(master_conn);
 
 	/*
 	 * Cancel any query that is still being executed, so i can insert the
@@ -1152,8 +1146,8 @@ do_master_failover(void)
 			  total_nodes, visible_nodes);
 
 	/*
-	 * am i on the group that should keep alive? if i see less than half of
+	 * Am I on the group that should keep alive? If I see less than half of
-	 * total_nodes then i should do nothing
+	 * total_nodes then I should do nothing
 	 */
 	if (visible_nodes < (total_nodes / 2.0))
 	{
@@ -1524,7 +1518,7 @@ do_master_failover(void)
 		my_local_conn = establish_db_connection(local_options.conninfo, true);
 
 		/* update node information to reflect new status */
-		if(update_node_record_set_upstream(new_master_conn, node_info.node_id, best_candidate.node_id) == false)
+		if (update_node_record_set_upstream(new_master_conn, local_options.cluster_name, node_info.node_id, best_candidate.node_id) == false)
 		{
 			appendPQExpBuffer(&event_details,
 							  _("Unable to update node record for node %i (following new upstream node %i)"),
@@ -1592,7 +1586,7 @@ do_upstream_standby_failover(t_node_info upstream_node)
 	 * Verify that we can still talk to the cluster master even though
 	 * node upstream is not available
 	 */
-	if (!check_connection(master_conn, "master"))
+	if (!check_connection(&master_conn, "master", NULL))
 	{
 		log_err(_("do_upstream_standby_failover(): Unable to connect to last known master node\n"));
 		return false;
@@ -1668,7 +1662,7 @@ do_upstream_standby_failover(t_node_info upstream_node)
 		terminate(ERR_BAD_CONFIG);
 	}
 
-	if(update_node_record_set_upstream(master_conn, node_info.node_id, upstream_node_id) == false)
+	if (update_node_record_set_upstream(master_conn, local_options.cluster_name, node_info.node_id, upstream_node_id) == false)
 	{
 		terminate(ERR_BAD_CONFIG);
 	}
@@ -1681,7 +1675,7 @@ do_upstream_standby_failover(t_node_info upstream_node)
 
 
 static bool
-check_connection(PGconn *conn, const char *type)
+check_connection(PGconn **conn, const char *type, const char *conninfo)
 {
 	int			connection_retries;
 
@@ -1692,7 +1686,16 @@ check_connection(PGconn *conn, const char *type)
 	 */
 	for (connection_retries = 0; connection_retries < local_options.reconnect_attempts; connection_retries++)
 	{
-		if (!is_pgup(conn, local_options.master_response_timeout))
+		if (*conn == NULL)
+		{
+			if (conninfo == NULL)
+			{
+				log_err("INTERNAL ERROR: *conn == NULL && conninfo == NULL");
+				terminate(ERR_INTERNAL);
+			}
+			*conn = establish_db_connection(conninfo, false);
+		}
+		if (!is_pgup(*conn, local_options.master_response_timeout))
 		{
 			log_warning(_("connection to %s has been lost, trying to recover... %i seconds before failover decision\n"),
 						type,
@@ -1710,9 +1713,9 @@ check_connection(PGconn *conn, const char *type)
 		}
 	}
 
-	if (!is_pgup(conn, local_options.master_response_timeout))
+	if (!is_pgup(*conn, local_options.master_response_timeout))
 	{
-		log_err(_("unable to reconnect to %s after %i seconds...\n"),
+		log_err(_("unable to reconnect to %s (timeout %i seconds)...\n"),
 				type,
 				local_options.master_response_timeout
 			);
@@ -1740,7 +1743,7 @@ set_local_node_failed(void)
 	int			active_master_node_id = NODE_NOT_FOUND;
 	char		master_conninfo[MAXLEN];
 
-	if (!check_connection(master_conn, "master"))
+	if (!check_connection(&master_conn, "master", NULL))
 	{
 		log_err(_("set_local_node_failed(): Unable to connect to last known master node\n"));
 		return false;
@@ -1966,17 +1969,21 @@ usage(void)
 void
 help(const char *progname)
 {
-	printf(_("Usage: %s [OPTIONS]\n"), progname);
+	printf(_("%s: replication management daemon for PostgreSQL\n"), progname);
-	printf(_("Replicator manager daemon for PostgreSQL.\n"));
+	printf(_("\n"));
-	printf(_("\nOptions:\n"));
+	printf(_("Usage:\n"));
-	printf(_("  --help                    show this help, then exit\n"));
+	printf(_("  %s [OPTIONS]\n"), progname);
-	printf(_("  --version                 output version information, then exit\n"));
+	printf(_("\n"));
+	printf(_("Options:\n"));
+	printf(_("  -?, --help                show this help, then exit\n"));
+	printf(_("  -V, --version             output version information, then exit\n"));
 	printf(_("  -v, --verbose             output verbose activity information\n"));
 	printf(_("  -m, --monitoring-history  track advance or lag of the replication in every standby in repl_monitor\n"));
 	printf(_("  -f, --config-file=PATH    path to the configuration file\n"));
 	printf(_("  -d, --daemonize           detach process from foreground\n"));
 	printf(_("  -p, --pid-file=PATH       write a PID file\n"));
-	printf(_("\n%s monitors a cluster of servers.\n"), progname);
+	printf(_("\n"));
+	printf(_("%s monitors a cluster of servers and optionally performs failover.\n"), progname);
 }
 
 
@@ -2219,23 +2226,12 @@ check_and_create_pid_file(const char *pid_file)
 t_node_info
 get_node_info(PGconn *conn, char *cluster, int node_id)
 {
-	char		sqlquery[QUERY_STR_LEN];
 	PGresult   *res;
 
 	t_node_info node_info = { NODE_NOT_FOUND, NO_UPSTREAM_NODE, "", InvalidXLogRecPtr, UNKNOWN, false, false};
 
-	sprintf(sqlquery,
+	res = get_node_record(conn, cluster, node_id);
-			"SELECT id, upstream_node_id, conninfo, type, slot_name, active "
-			"  FROM %s.repl_nodes "
-			" WHERE cluster = '%s' "
-			"   AND id = %i",
-			get_repmgr_schema_quoted(conn),
-			local_options.cluster_name,
-			node_id);
 
-	log_debug("get_node_info(): %s\n", sqlquery);
-
-	res = PQexec(my_local_conn, sqlquery);
 	if (PQresultStatus(res) != PGRES_TUPLES_OK)
 	{
 		PQExpBufferData errmsg;
@@ -2299,37 +2295,3 @@ parse_node_type(const char *type)
 
 	return UNKNOWN;
 }
-
-
-static bool
-update_node_record_set_upstream(PGconn *conn, int this_node_id, int new_upstream_node_id)
-{
-	PGresult   *res;
-	char		sqlquery[QUERY_STR_LEN];
-
-	log_debug(_("update_node_record_set_upstream(): Updating node %i's upstream node to %i\n"), this_node_id, new_upstream_node_id);
-
-	sqlquery_snprintf(sqlquery,
-					  "  UPDATE %s.repl_nodes "
-					  "     SET upstream_node_id = %i "
-					  "   WHERE cluster = '%s' "
-					  "     AND id = %i ",
-					  get_repmgr_schema_quoted(conn),
-					  new_upstream_node_id,
-					  local_options.cluster_name,
-					  this_node_id);
-	res = PQexec(conn, sqlquery);
-
-	if (PQresultStatus(res) != PGRES_COMMAND_OK)
-	{
-		log_err(_("Unable to set new upstream node id: %s\n"),
-				PQerrorMessage(conn));
-		PQclear(res);
-
-		return false;
-	}
-
-	PQclear(res);
-
-	return true;
-}

version.h

@@ -1,6 +1,6 @@
 #ifndef _VERSION_H_
 #define _VERSION_H_
 
-#define REPMGR_VERSION "3.0.1"
+#define REPMGR_VERSION "3.0.2"
 
 #endif