Postgres: Purpose of archiving in master?

Question

I am reading tutorials on how to setup streaming replication in Postgres from master to slave.

Some tutorials recommend setting up archiving in the master (and restoring in slave) . In the master, this needs to be set in the conf file:

wal_level = hot_standby
archive_mode = on
archive_command = 'some rsync command to rsync logfile segments from master to slave'

What is the purpose of this archiving? When is it not necessary when setting up streaming replication?

Answer 1

WAL archiving is useful when you're running streaming replication, because there's a limit to how much WAL the master will retain.

If you don't archive WAL, and the replica gets s far behind that the master has discarded WAL it still needs, it cannot recover and must be replaced with a fresh base backup from the master.

It's also useful for PITR for disaster recovery purposes.

• Henley Wing Chiu: is it too late to setup archiving when I already had the replication from master to slave running for a long time? If I can, should I restart the slave first?
• Craig Ringer: @HenleyChiu You can start WAL archiving at any time. You'll need to restart the replica to set a recovery.conf entry for the restore_command. The master may also need restarting to set an archive_command.
• Henley Wing Chiu: should I restart the slave first then the master?
• Craig Ringer: @HenleyChiu WAL archiving is asynchronous, so it doesn't really matter, does it?