Tomonari Katsumata wrote:
> Hi Alvaro,
>
> >> do you have another patch?
> >
> > Sorry, yes. This is the complete series of patches, although I have not
> > tried them against 9.3 yet.
> >
> Thank you.
> I could apply them against 93_STABLE and build it.
> (commitID: d32e8387bac15ad353029c259e832f78d456c7a8)
>
> deadlock did not occur with my test case.
> It seems these patches resolve my problem.
>
> If it doesn't have any other problems,
> would you commit them for next release(9.3.2)?
Yes, it's already committed for 9.3.2. There are a bunch of other fixes
related to multixact which I intend to commit soon too.
> FYI:
> These patch doesn't resolve bug#8470 yet.
> I could reproduce the problem on patched PostgreSQL.
Yeah, I tested that too, but that problem is not related. I also tested
it with the fix to another performance problem which I posted as
http://www.postgresql.org/message-id/20131127222435.GC5513@eldon.alvh.no-ip.org
but that's not it either. It really requires a completely separate
optimization.
Thanks,
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services