Andres Freund <and...@anarazel.de> writes:
> On February 13, 2019 1:16:07 PM GMT+01:00, Peter Eisentraut
> <peter.eisentr...@2ndquadrant.com> wrote:
>> One idea to address this is to slow down WAL-generating maintenance
>> operations. This is similar to the vacuum delay. Where the vacuum
>> delay counts notional I/O cost before sleeping, here we would count how
>> much WAL has been generated and sleep after some amount.
> Interesting idea, not yet quite sure what to think. But I don't think the way
> you did it is acceptable - we can't just delay while holding buffer locks, in
> critical sections, while not interruptible.
Yeah. Maybe it could be done in a less invasive way by just having the
WAL code keep a running sum of how much WAL this process has created,
and then letting the existing vacuum-delay infrastructure use that as
one of its how-much-IO-have-I-done inputs.
Not sure if that makes the tuning problem easier or harder, but
it seems reasonable on its face to count WAL emission as I/O.
regards, tom lane
