Operation when the standby server is stopped
Before performing maintenance for the primary server instance when the standby server has been stopped, comment out the synchronous_standby_names parameter in the postgresql.conf file of the primary server, and then execute the pg_ctl command in reload mode.
If this operation is not performed, operations performed on the primary server for the instance will remain in a wait state.
See
Refer to "pg_ctl" in Reference for information on the command.
Standby server downtime
If you specified the synchronous_standby_names parameter of the postgresql.conf file and then the standby server instance is stopped, consider the points below.
The replication_timeout parameter in the postgresql.conf file
If the standby server is stopped after the timeout set in this parameter was exceeded, an error stating that the transaction log could not be received may be output to the standby server system log, and all transaction logs that should be transferred to the standby server may be lost.
The wal_keep_segments parameter in the postgresql.conf file
If a transaction log that exceeds the value set in this parameter was generated while the standby server was stopped, the transaction log may be deleted.
Additionally, setting this parameter requires consideration regarding stabilization of the database multiplexing mode. Refer to "6.9.1 Tuning to Stabilize the Database Multiplexing Mode" for details.
Note
The standby server must be rebuilt in both cases above.
Take the action advised in the recovery operation that starts from "8.1.1.1.3 Identify cause of error and perform recovery" through to "8.1.1.2 Rebuild the Standby Server".
