Michal Taborsky <> writes:
> I've read the discussion in "Trying to minimize the impact of
> checkpoints" thread and I get it, that there is nothing I can do about
> it. Well, we'll have to live with that, at least until 7.5.
You could experiment with the checkpoint interval (checkpoint_timeout).
A shorter interval will mean more total I/O (the same page will get
written out more often) but it should reduce the amount of I/O done by
any one checkpoint. You might find that the extra overhead is worth it
to reduce the spikes.
But 7.5 should provide a much better answer, yes.
regards, tom lane