On Sun, Jul 7, 2019 at 12:15 PM Thorsten Schöning <tschoening@xxxxxxxxxx> wrote:
From my understanding, without actually archving WAL segments, the
WAL would simply grow until archiving gets enabled, without any
influence on if the backup is consistent or not.
This is true if you set archive_command to the empty string, or to a command that returns a failure code (like the linux command 'false'). But if you turn archive_mode to off, or if you use an archive_command that always returns success despite not archiving (like the linux command 'true') then the WAL would not be retained but would be lost.
So you would have to go through a complex cycle of having it return success without archiving, then just before the backup change it to return failure, then change it to actually archive and return success, then once all needed WAL is archived change once more to return success without archiving.
That is a lot of complexity. What does it get you? The normal way has the archive happen at the same time as the base backup, and I guess that that network traffic could delay the basebackup itself, which then means you have to archive the WAL that was generated during that delay. That is something, but it doesn't seem like much to justify the complexity.
Cheers,
Jeff
