repmgrd: include node name in log output in more places

Still a few places where only the node ID was reported, but it's always
useful to have the node name as well.
This commit is contained in:
Ian Barwick
2020-05-11 16:55:31 +09:00
parent 1b5ad743b5
commit fd52df0fab

View File

@@ -1715,7 +1715,8 @@ monitor_streaming_standby(void)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("reconnected to upstream node %i after %i seconds, resuming monitoring"), _("reconnected to upstream node \"%s\" (ID: %i) after %i seconds, resuming monitoring"),
upstream_node_info.node_name,
upstream_node_info.node_id, upstream_node_info.node_id,
degraded_monitoring_elapsed); degraded_monitoring_elapsed);
@@ -2036,7 +2037,8 @@ loop:
if (last_known_upstream_node_id != local_node_info.upstream_node_id) if (last_known_upstream_node_id != local_node_info.upstream_node_id)
{ {
log_notice(_("local node %i upstream appears to have changed, restarting monitoring"), log_notice(_("upstream for local node \"%s\" (ID: %i) appears to have changed, restarting monitoring"),
local_node_info.node_name,
local_node_info.node_id); local_node_info.node_id);
log_detail(_("currently monitoring upstream %i; new upstream is %i"), log_detail(_("currently monitoring upstream %i; new upstream is %i"),
last_known_upstream_node_id, last_known_upstream_node_id,
@@ -2050,7 +2052,8 @@ loop:
*/ */
if (local_node_info.type != STANDBY) if (local_node_info.type != STANDBY)
{ {
log_notice(_("local node %i is no longer a standby, restarting monitoring"), log_notice(_("local node \"%s\" (ID: %i) is no longer a standby, restarting monitoring"),
local_node_info.node_name,
local_node_info.node_id); local_node_info.node_id);
close_connection(&upstream_conn); close_connection(&upstream_conn);
return; return;
@@ -2144,7 +2147,8 @@ loop:
if (last_known_upstream_node_id != local_node_info.upstream_node_id) if (last_known_upstream_node_id != local_node_info.upstream_node_id)
{ {
log_notice(_("local node %i's upstream appears to have changed, restarting monitoring"), log_notice(_("local node \"%s\" (ID: %i)'s upstream appears to have changed, restarting monitoring"),
local_node_info.node_name,
local_node_info.node_id); local_node_info.node_id);
log_detail(_("currently monitoring upstream %i; new upstream is %i"), log_detail(_("currently monitoring upstream %i; new upstream is %i"),
last_known_upstream_node_id, last_known_upstream_node_id,
@@ -2397,7 +2401,8 @@ monitor_streaming_witness(void)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("reconnected to upstream node %i after %i seconds, resuming monitoring"), _("reconnected to upstream node \"%s\" (ID: %i) after %i seconds, resuming monitoring"),
upstream_node_info.node_name,
upstream_node_info.node_id, upstream_node_info.node_id,
degraded_monitoring_elapsed); degraded_monitoring_elapsed);
@@ -2407,7 +2412,8 @@ monitor_streaming_witness(void)
if (get_recovery_type(primary_conn) != RECTYPE_PRIMARY) if (get_recovery_type(primary_conn) != RECTYPE_PRIMARY)
{ {
log_notice(_("current upstream node \"%s\" (ID: %i) is not primary, restarting monitoring"), log_notice(_("current upstream node \"%s\" (ID: %i) is not primary, restarting monitoring"),
upstream_node_info.node_name, upstream_node_info.node_id); upstream_node_info.node_name,
upstream_node_info.node_id);
PQfinish(primary_conn); PQfinish(primary_conn);
primary_conn = NULL; primary_conn = NULL;
termPQExpBuffer(&event_details); termPQExpBuffer(&event_details);
@@ -2725,14 +2731,16 @@ do_primary_failover(void)
if (sibling_wal_receiver_pid == UNKNOWN_PID) if (sibling_wal_receiver_pid == UNKNOWN_PID)
{ {
log_warning(_("unable to query WAL receiver PID on node %i"), log_warning(_("unable to query WAL receiver PID on node \"%s\" (ID: %i)"),
cell->node_info->node_name,
cell->node_info->node_id); cell->node_info->node_id);
} }
else if (sibling_wal_receiver_pid > 0) else if (sibling_wal_receiver_pid > 0)
{ {
log_info(_("WAL receiver PID on node %i is %i"), log_info(_("WAL receiver PID on node node \"%s\" (ID: %i) is %i"),
cell->node_info->node_name,
cell->node_info->node_id, cell->node_info->node_id,
sibling_wal_receiver_pid); (int)sibling_wal_receiver_pid);
sibling_node_wal_receiver_connected = true; sibling_node_wal_receiver_connected = true;
} }
} }
@@ -2895,7 +2903,8 @@ do_primary_failover(void)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("node %i is in manual failover mode and is now disconnected from streaming replication"), _("node \"%s\" (ID: %i) is in manual failover mode and is now disconnected from streaming replication"),
local_node_info.node_name,
local_node_info.node_id); local_node_info.node_id);
new_primary_conn = establish_db_connection(new_primary.conninfo, false); new_primary_conn = establish_db_connection(new_primary.conninfo, false);
@@ -3299,7 +3308,8 @@ do_upstream_standby_failover(void)
if (PQstatus(local_conn) != CONNECTION_OK) if (PQstatus(local_conn) != CONNECTION_OK)
{ {
log_error(_("unable to reconnect to local node %i"), log_error(_("unable to reconnect to local node \"%s\" (ID: %i)"),
local_node_info.node_name,
local_node_info.node_id); local_node_info.node_id);
return FAILOVER_STATE_FOLLOW_FAIL; return FAILOVER_STATE_FOLLOW_FAIL;
} }
@@ -3334,7 +3344,8 @@ do_upstream_standby_failover(void)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("unable to set node %i's new upstream ID to %i"), _("unable to set node \"%s\" (ID: %i)'s new upstream ID to %i"),
local_node_info.node_name,
local_node_info.node_id, local_node_info.node_id,
primary_node_info.node_id); primary_node_info.node_id);
@@ -3372,8 +3383,10 @@ do_upstream_standby_failover(void)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("node %i is now following primary node %i"), _("node \"%s\" (ID: %i) is now following primary node \"%s\" (ID: %i)"),
local_node_info.node_name,
local_node_info.node_id, local_node_info.node_id,
primary_node_info.node_name,
primary_node_info.node_id); primary_node_info.node_id);
log_notice("%s", event_details.data); log_notice("%s", event_details.data);
@@ -3527,8 +3540,10 @@ promote_self(void)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("node %i promoted to primary; old primary %i marked as failed"), _("node \"%s\" (ID: %i) promoted to primary; old primary \"%s\" (ID: %i) marked as failed"),
local_node_info.node_name,
local_node_info.node_id, local_node_info.node_id,
failed_primary.node_name,
failed_primary.node_id); failed_primary.node_id);
/* local_conn is now the primary connection */ /* local_conn is now the primary connection */
@@ -3814,7 +3829,8 @@ follow_new_primary(int new_primary_id)
if (PQstatus(local_conn) != CONNECTION_OK) if (PQstatus(local_conn) != CONNECTION_OK)
{ {
log_error(_("unable to reconnect to local node %i"), log_error(_("unable to reconnect to local node \"%s\" (ID: %i)"),
local_node_info.node_name,
local_node_info.node_id); local_node_info.node_id);
return FAILOVER_STATE_FOLLOW_FAIL; return FAILOVER_STATE_FOLLOW_FAIL;
} }
@@ -3828,8 +3844,10 @@ follow_new_primary(int new_primary_id)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("node %i now following new upstream node %i"), _("node \"%s\" (ID: %i) now following new upstream node \"%s\" (ID: %i)"),
local_node_info.node_name,
local_node_info.node_id, local_node_info.node_id,
upstream_node_info.node_name,
upstream_node_info.node_id); upstream_node_info.node_id);
log_notice("%s", event_details.data); log_notice("%s", event_details.data);
@@ -3924,7 +3942,7 @@ witness_follow_new_primary(int new_primary_id)
record_status = get_node_record(upstream_conn, local_node_info.node_id, &local_node_info); record_status = get_node_record(upstream_conn, local_node_info.node_id, &local_node_info);
if (record_status != RECORD_FOUND) if (record_status != RECORD_FOUND)
{ {
log_error(_("unable to retrieve metadata record found for node %i"), log_error(_("unable to retrieve metadata record for node %i"),
local_node_info.node_id); local_node_info.node_id);
return FAILOVER_STATE_FOLLOW_FAIL; return FAILOVER_STATE_FOLLOW_FAIL;
} }
@@ -3934,8 +3952,10 @@ witness_follow_new_primary(int new_primary_id)
initPQExpBuffer(&event_details); initPQExpBuffer(&event_details);
appendPQExpBuffer(&event_details, appendPQExpBuffer(&event_details,
_("witness node %i now following new primary node %i"), _("witness node \"%s\" (ID: %i) now following new primary node \"%s\" (ID: %i)"),
local_node_info.node_name,
local_node_info.node_id, local_node_info.node_id,
upstream_node_info.node_name,
upstream_node_info.node_id); upstream_node_info.node_id);
log_notice("%s", event_details.data); log_notice("%s", event_details.data);
@@ -4301,7 +4321,8 @@ do_election(NodeInfoList *sibling_nodes, int *new_primary_id)
{ {
if (sibling_replication_info.upstream_node_id != upstream_node_info.node_id) if (sibling_replication_info.upstream_node_id != upstream_node_info.node_id)
{ {
log_warning(_("assumed sibling node %i monitoring different upstream node %i"), log_warning(_("assumed sibling node \"%s\" (ID: %i) monitoring different upstream node %i"),
cell->node_info->node_name,
cell->node_info->node_id, cell->node_info->node_id,
sibling_replication_info.upstream_node_id); sibling_replication_info.upstream_node_id);
@@ -4341,7 +4362,8 @@ do_election(NodeInfoList *sibling_nodes, int *new_primary_id)
/* don't check 0-priority nodes */ /* don't check 0-priority nodes */
if (cell->node_info->priority <= 0) if (cell->node_info->priority <= 0)
{ {
log_info(_("node %i has priority of %i, skipping"), log_info(_("node \"%s\" (ID: %i) has priority of %i, skipping"),
cell->node_info->node_name,
cell->node_info->node_id, cell->node_info->node_id,
cell->node_info->priority); cell->node_info->priority);
continue; continue;
@@ -5176,11 +5198,16 @@ try_primary_reconnect(PGconn **conn, PGconn *local_conn, t_node_info *node_info)
for (i = 0; i < max_attempts; i++) for (i = 0; i < max_attempts; i++)
{ {
log_info(_("checking state of node %i, %i of %i attempts"), log_info(_("checking state of node \"%s\" (ID: %i), %i of %i attempts"),
node_info->node_id, i + 1, max_attempts); node_info->node_name,
node_info->node_id,
i + 1, max_attempts);
if (is_server_available_params(&conninfo_params) == true) if (is_server_available_params(&conninfo_params) == true)
{ {
log_notice(_("node %i has recovered, reconnecting"), node_info->node_id); log_notice(_("node \"%s\" (ID: %i) has recovered, reconnecting"),
node_info->node_name,
node_info->node_id);
/* /*
* Note: we could also handle the case where node is pingable but * Note: we could also handle the case where node is pingable but
@@ -5193,7 +5220,9 @@ try_primary_reconnect(PGconn **conn, PGconn *local_conn, t_node_info *node_info)
{ {
free_conninfo_params(&conninfo_params); free_conninfo_params(&conninfo_params);
log_info(_("connection to node %i succeeded"), node_info->node_id); log_info(_("connection to node \"%s\" (ID: %i) succeeded"),
node_info->node_name,
node_info->node_id);
if (PQstatus(*conn) == CONNECTION_BAD) if (PQstatus(*conn) == CONNECTION_BAD)
{ {
@@ -5257,7 +5286,8 @@ try_primary_reconnect(PGconn **conn, PGconn *local_conn, t_node_info *node_info)
} }
} }
log_warning(_("unable to reconnect to node %i after %i attempts"), log_warning(_("unable to reconnect to node \"%s\" (ID: %i) after %i attempts"),
node_info->node_name,
node_info->node_id, node_info->node_id,
max_attempts); max_attempts);