Hello
I run two PostgreSQL servers in a master-slave setup and set
wal_keep_segments=1000 on the master to allow long downtimes on the slave.
Meanwhile the disk got fuller than I estimated and I changed the config
to wal_keep_segments=500 and restarted the server afterwards.
Yet, the number of WAL segments in my archive dir was still at 1000!
I tried a random mix of VACUUM, CHECKPOINT, pg_start_backup(),
pg_stop_backup(), server restarts, googling and just plain waiting
but the number of archived WAL segments just increased to 1018.
Although I understand that I may safely delete those WAL files that
are older than the last .backup file I wonder if that shouldn't be
happen automatically. Why doesn't it?
Relevant non-default settings from the master with version 9.1.6:
wal_level = hot_standby
archive_mode = on
archive_command = 'test ! -f /srv/postgresql-archivedir/%f && cp %p /srv/postgresql-archivedir/%f'
wal_keep_segments = 500
hot_standby = on
bye,
-christian-