Re: [HACKERS] WAL archiving idle database

Jeff Davis <pgsql(at)j-davis(dot)com> writes:
> I would think if the current location does not end in all zeros, you
> should expect a new WAL segment to be archived soon. Although this
> assumes that an idle database would not advance that location at all,
> and I'm still trying to understand Tom's proposal well enough to know
> whether that would be true or not.

With my proposal, after the last activity, you'd get a checkpoint, and
then at the next archive_timeout we'd advance the pointer to a segment
boundary and archive the old segment, and then nothing more would happen
until the next WAL-loggable update. So yeah, the master's
pg_current_xlog_location could be expected to sit at a segment boundary
while it was idle.

regards, tom lane