Jeff Davis <pgsql@j-davis.com> writes:
> Keep in mind that even in the current system, your configuration is
> variable based on the checkpoint_timeout setting.
Yeah, and he has to keep this less than archive_timeout in order for
it to work the way he wants, which is probably not good for performance.
(Sane settings of checkpoint_timeout are probably higher, not lower,
than what people are likely to use for archive_timeout.)
I think my recommendation to Kevin would be to force some trivial
transaction to occur a little before each expected archive_timeout,
so that there will be something to be archived. This would have the
additional advantage that the monitor is checking that the database is
actually responding to queries, whereas just noting that it's spitting
out WAL files doesn't really prove that --- especially not if mere
no-op checkpoints can cause WAL files to be emitted.
regards, tom lane