Re: Backport of fsync queue compaction

On Tue, Jun 19, 2012 at 5:39 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Tue, Jun 19, 2012 at 5:33 PM, Greg Smith <greg(at)2ndquadrant(dot)com> wrote:
>> In January of 2011 Robert committed 7f242d880b5b5d9642675517466d31373961cf98
>> to try and compact the fsync queue when clients find it full.  There's no
>> visible behavior change, just a substantial performance boost possible in
>> the rare but extremely bad situations where the background writer stops
>> doing fsync absorption.  I've been running that in production at multiple
>> locations since practically the day it hit this mailing list, with backports
>> all the way to 8.3 being common (and straightforward to construct).  I've
>> never seen a hint of a problem with this new code.
>
> I've been in favor of back-porting this for a while, so you'll get no
> argument from me.
>
> Anyone disagree?

Hearing no disagreement, I went ahead and did this. I didn't take
Greg Smith's suggestion of adding a log message when/if the fsync
compaction logic fails to make any headway, because (1) the proposed
message didn't follow message style guidelines and I couldn't think of
a better one that did and (2) I'm not sure it's worth creating extra
translation work in the back-branches for such a marginal case. We
can revisit this if people feel strongly about it.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company