Howdy, I had an instance where a replica fell out of sync with the master. Now it's in in a state where it's unable to catch up because the master has already removed the WAL segment. (logs) Mar 2 23:10:13 db13 postgres[11099]: [3-1] user=,db=,host= LOG: streaming replication successfully connected to primary Mar 2 23:10:13 db13 postgres[11099]: [4-1] user=,db=,host= FATAL: could not receive data from WAL stream: FATAL: requested WAL segment 000000060000047C0000001F has already been removed I was under the impression that when you setup streaming replication if you specify a restore command like : restore_command = 'cp /arch/%f %p' Then even if the slave falls out of sync, and the master removes the WAL segment, as long as you can still retrieve the WAL files, then it can bring itself back into sync. But that doesn't seem to be happening. The restore_command is working # Slave's $PGDATA/pg_xlog/ -rw------- 1 postgres postgres 16777216 Mar 2 21:29 000000060000047C0000001F -rwx------ 1 postgres postgres 16777216 Mar 2 23:13 RECOVERYXLOG I'm on PG 9.2.7, which i know is old, but I'm upgrading shortly. recovery.conf: standby_mode = 'on' primary_conninfo = 'host=pgmaster port=5432' restore_command = 'cp /arch/%f %p' relevant info from postgresql.conf: wal_level = hot_standby max_wal_senders = 5 wal_keep_segments = 32 archive_mode = on hot_standby = on hot_standby_feedback = true I know that to avoid this entirely I need to set wal_keep_segments higher, although in this particular case it wouldn't have mattered because a rogue program slammed the database and basically 32/64/128 WAL segments went by in a short span of time. However, I really thought that as long as PG could get the archived logs i'd be able to recover. Was I wrong with that assertion or did i just run into a bug? Thanks -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
