Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> I think the point is that a totally idle database should not
>> continue to emit WAL, not even at a slow rate. There are also
>> power-consumption objections to allowing the checkpoint process
>> to fire up to no purpose.
>
> Hmm, OK. I still think it's a little funny to say that
> checkpoint_timeout will force a checkpoint every N minutes except
> when it doesn't, but maybe there's no real harm in that as long as
> we document it properly.
What will be the best way to determine, from looking only at a
standby, whether replication is healthy and progressing? Going back
to the 8.1 warm standby days we have run pg_controldata and compared
the "Time of latest checkpoint" to current time; I'm hoping there's
a better way now, so that we can drop that kludge. If not, I would
like a way to kick out a checkpoint from the master at some finite
and configurable interval for monitoring purposes. I'm all for
having a nice, sharp fillet knife, but if that's not available,
please don't take away this rock I chipped at until it had an
edge.
Most likely there's something there which I've missed, but it's
really nice to be able to tell the difference between a quiet master
and broken replication.
-Kevin