Em sex, 21 de set de 2018 às 09:19, Prince Pathria <prince.pathria@xxxxxxxxxxx> escreveu: > > Hi, > > On postgresql10, I'm facing issue of "requested WAL segment 000000010000007E0000003F has already been removed" a lot. I resync slaves with pg_basebackup and still some days after the same thing happens with same or other replica node. > > I have kept > > wal_keep_segment = 1024 > wal_buffers = 16MB > min_wal_size = 1GB > max_wal_size = 2GB > archive_mode = on > archive_command = cp %p /var/lib/pgsql/10/archive/%f > wal_level = replica > synchronous_commit = on > synchronous_standby_names = 'pgslave01' > > Can I keep wal segments for some duration(eg 24 hours) ? Think about replication slots to deal with it: https://www.postgresql.org/docs/current/static/warm-standby.html#STREAMING-REPLICATION-SLOTS > I'm using two replicas with one in sync and other in potential state.Is one synchronous_standby_names fine? It seems ok, you'll have one synchronous replica. > Is there any option to remove archive wal after 30 days or 50GB? You have to keep wal that corresponds to your oldest basebackup. Consider using one of the backup solutions like Barman or pgbackrest. Barman works with basebackup and replication slots and you can set a retention policy for your backups, it handles wal keeping for you. > And how archiving the wal logs help? The missing segment is "000000010000007E0000003F" is in archive dir but how can I make it in use? You have to setup a restore_command to retrieve it from your archive: https://www.postgresql.org/docs/10/static/archive-recovery-settings.html#RESTORE-COMMAND
