The following error messages and conditions may be encountered while using the recovery kit.
|Unable to protect PostgreSQL database using the same port as another LK protected PostgreSQL database.|| Verify the version of PostgreSQL includes a postgresql.conf file. In the postgresql.conf file, set the entry listen_address= to the IP address to be used with the database instance.
Note: The format of the listen_address= in the postgresql.conf file is important as syntax errors can result in a failure to start the database server.
|Unable to perform a manual switchover of version 8.X when clients are connected.||The default (smart) shutdown option fails to disconnect clients on a switchover. If shutdown continues to fail with connected clients, verify that the LKPGSQL_SDIRS tunable is not set. If the problem persists, set the LifeKeeper tunable LKPGSQL_IDIRS.|
|Unable to connect from a remote client to the database server.||To enable remote host login for PostgreSQL, refer to the PostgreSQL Administration Guide on configuring the pg_hba.conf file.|
|psql: connectDBStart() — connect() failed: No such file or directory. Is the postmaster running at ‘localhost’ and accepting connections on Unix socket ‘<port>’?”||Verify that the socket file exists and the instance is currently running. If the socket file resides in /tmp, it may have been removed by a cron job that cleans up the /tmp directory. Take the resource out of service and back in service. Then modify the cron job to leave PostgreSQL socket files.|
|PostgreSQL resource hierarchy fails to come in service but the database is running.||The database may have failed to respond to the LifeKeeper client request within the specified interval. Adjust the tunable LKPGSQL_CONN_RETRIES in /etc/default/LifeKeeper to increase the number of seconds allowed for the recovery and restart of the PostgreSQL database instance.|
|PostgreSQL resource hierarchy fails local recovery following a postmaster crash with active client connections.||When a large number of active clients are connected to PostgreSQL, the database may be unable to properly restart until the client connections have terminated. In this scenario, it may be best to force client connections to terminate so that local recovery will be successful. The variable LKPGSQL_DISCONNECT_CLIENT can be set in /etc/default/LifeKeeper to control the behavior of the PostgreSQL resource hierarchy in this scenario. When the value is set to 1(true), client processes will be sent a SIGTERM signal to force them to disconnect from the database. This action will only be taken if the postmaster process is not running during local recovery.|
|Unable to connect to database with error “WARNING: password file “/home/<osuser>/.pgpass” has world or group read access”||The .pgpass file permissions should be u=rw(0600). Change the permissions and owner of the .pgpass file.|
|FATAL: syntax error in file “/<datadir>/postgresql.conf” line 50, near token “.17”||The postgresql.conf file listen_address= entry does not contain proper quoting. Verify entries are valid and the entry is enclosed in proper quotes.|
Thanks for your feedback.
Post your comment on this topic.