ocf_heartbeat_pgsql - Man Page

pgsql [start | stop | status | monitor | promote | demote | meta-data | validate-all]

Resource script for PostgreSQL. It manages a PostgreSQL as an HA resource.

pgctl

Path to pg_ctl command.

(optional, string, default "/usr/bin/pg_ctl")

start_opt

Start options (-o start_opt in pg_ctl). "-i -p 5432" for example.

(optional, string, no default)

ctl_opt

Additional pg_ctl options (-w, -W etc..).

(optional, string, no default)

psql

Path to psql command.

(optional, string, default "/usr/bin/psql")

pgdata

Path to PostgreSQL data directory.

(optional, string, default "/var/lib/pgsql/data")

pgdba

User that owns PostgreSQL.

(optional, string, default "postgres")

pghost

Hostname/IP address where PostgreSQL is listening

(optional, string, no default)

pgport

Port where PostgreSQL is listening

(optional, integer, default 5432)

pglibs

Custom location of the Postgres libraries. If not set, the standard location will be used.

(optional, string, default "/usr/lib")

monitor_user

PostgreSQL user that pgsql RA will user for monitor operations. If it's not set pgdba user will be used.

(optional, string, no default)

monitor_password

Password for monitor user.

(optional, string, no default)

monitor_sql

SQL script that will be used for monitor operations.

(optional, string, default "select now();")

config

Path to the PostgreSQL configuration file for the instance.

(optional, string, default "/var/lib/pgsql/data/postgresql.conf")

pgdb

Database that will be used for monitoring.

(optional, string, default "template1")

logfile

Path to PostgreSQL server log output file.

(optional, string, default "/dev/null")

socketdir

Unix socket directory for PostgreSQL.

If you use PostgreSQL 9.3 or higher and define unix_socket_directories in the postgresql.conf, then you must set socketdir to determine which directory is used for psql command.

(optional, string, no default)

stop_escalate

Number of seconds to wait for stop (using -m fast) before resorting to -m immediate

(optional, integer, default 90)

rep_mode

Replication mode may be set to "async" or "sync" or "slave". They require PostgreSQL 9.1 or later. Once set, "async" and "sync" require node_list, master_ip, and restore_command parameters,as well as configuring PostgreSQL for replication (in postgresql.conf and pg_hba.conf).

"slave" means that RA only makes recovery.conf before starting to connect to primary which is running somewhere. It doesn't need master/slave setting. It requires master_ip restore_command parameters.

(optional, string, default "none")

node_list

All node names. Please separate each node name with a space. This is optional for replication. Defaults to all nodes in the cluster

(optional, string, no default)

restore_command

restore_command for recovery.conf. This is required for replication.

(optional, string, no default)

archive_cleanup_command

archive_cleanup_command for recovery.conf. This is used for replication and is optional.

(optional, string, no default)

recovery_end_command

recovery_end_command for recovery.conf. This is used for replication and is optional.

(optional, string, no default)

master_ip

Master's floating IP address to be connected from hot standby. This parameter is used for "primary_conninfo" in recovery.conf. This is required for replication.

(optional, string, no default)

repuser

User used to connect to the master server. This parameter is used for "primary_conninfo" in recovery.conf. This is required for replication.

(optional, string, default "postgres")

primary_conninfo_opt

primary_conninfo options of recovery.conf except host, port, user and application_name. This is optional for replication.

(optional, string, no default)

restart_on_promote

If this is true, RA deletes recovery.conf and restarts PostgreSQL on promote to keep Timeline ID. It probably makes fail-over slower. It's recommended to set on-fail of promote up as fence. This is optional for replication.

(optional, boolean, default false)

replication_slot_name

Set this option when using replication slots. Can only use lower case letters, numbers and underscore for replication_slot_name.

The replication slots would be created for each node, with the name adding the node name as postfix. For example, replication_slot_name is "sample" and 2 slaves which are "node1" and "node2" connect to their slots, the slots names are "sample_node1" and "sample_node2". If the node name contains a upper case letter, hyphen and dot, those characters will be converted to a lower case letter or an underscore. For example, Node-1.example.com to node_1_example_com.

pgsql RA doesn't monitor and delete the replication slot. When the slave node has been disconnected in failure or the like, execute one of the following manually. Otherwise it may eventually cause a disk full because the master node will continue to accumulate the unsent WAL. 1. recover and reconnect the slave node to the master node as soon as possible. 2. delete the slot on the master node by following psql command. $ select pg_drop_replication_slot('replication_slot_name');

(optional, string, no default)

tmpdir

Path to temporary directory. This is optional for replication.

(optional, string, default "/var/lib/pgsql/tmp")

xlog_check_count

Number of checks of xlog on monitor before promote. This is optional for replication.

Note: For backward compatibility, the terms are unified with PostgreSQL 9. If you are using PostgreSQL 10 or later, replace "xlog" with "wal". Likewise, replacing "location" with "lsn".

(optional, integer, default 3)

crm_attr_timeout

The timeout of crm_attribute forever update command. Default value is 5 seconds. This is optional for replication.

(optional, integer, default 5)

stop_escalate_in_slave

Number of seconds to wait for stop (using -m fast) before resorting to -m immediate in slave state. This is optional for replication.

(optional, integer, default 90)

check_wal_receiver

If this is true, RA checks wal_receiver process on monitor and notifies its status using "(resource name)-receiver-status" attribute. It's useful for checking whether PostgreSQL (hot standby) connects to primary. The attribute shows status as "normal" or "normal (master)" or "ERROR". Note that if you configure PostgreSQL as master/slave resource, then wal receiver is not running in the master and the attribute shows status as "normal (master)" consistently because it is normal status.

(optional, boolean, default false)

This resource agent supports the following actions (operations):

start

Starts the resource. Suggested minimum timeout: 120s.

stop

Stops the resource. Suggested minimum timeout: 120s.

status

Performs a status check. Suggested minimum timeout: 60s.

monitor

Performs a detailed status check. Suggested minimum timeout: 30s. Suggested interval: 30s.

monitor (Promoted role)

Performs a detailed status check. Suggested minimum timeout: 30s. Suggested interval: 29s.

promote

Promotes the resource to the Master role. Suggested minimum timeout: 120s.

demote

Demotes the resource to the Slave role. Suggested minimum timeout: 120s.

notify

Suggested minimum timeout: 90s.

meta-data

Retrieves resource agent metadata (internal use only). Suggested minimum timeout: 5s.

validate-all

Performs a validation of the resource configuration. Suggested minimum timeout: 5s.

methods

Suggested minimum timeout: 5s.

The following is an example configuration for a pgsql resource using the crm(8) shell:

primitive p_pgsql ocf:heartbeat:pgsql \
  op monitor depth="0" timeout="30s" interval="30s" \
  op monitor depth="0" timeout="30s" interval="29s" role="Promoted"

ms ms_pgsql p_pgsql \
  meta notify="true" interleave="true"

The following is an example configuration for a pgsql resource using pcs(8)

pcs resource create p_pgsql ocf:heartbeat:pgsql \
  op monitor OCF_CHECK_LEVEL="0" timeout="30s" interval="30s" \
  op monitor OCF_CHECK_LEVEL="0" timeout="30s" interval="29s" role="Promoted" promotable

http://clusterlabs.org/

ClusterLabs contributors (see the resource agent source for information about individual authors)