Horton

Re: hung backends stuck in spinlock heavy endless loop

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Merlin Moncure <mmoncure@gmail.com>
Cc: Peter Geoghegan <pg@heroku.com>, Andres Freund <andres@2ndquadrant.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2015-01-14T00:05:10Z
Lists: pgsql-hackers 
Merlin Moncure <mmoncure@gmail.com> writes:
> On Tue, Jan 13, 2015 at 5:54 PM, Peter Geoghegan <pg@heroku.com> wrote:
>> In case it isn't clear, I think that the proximate cause here may well
>> be either one (or both) of commits
>> efada2b8e920adfdf7418862e939925d2acd1b89 and/or
>> 40dae7ec537c5619fc93ad602c62f37be786d161. Probably the latter. I think
>> that the profile is roughly consistent with that, although I may well
>> be wrong.

> I'm out of time for the day, but I'm fairly confident I can reproduce.
> I'll see if I can reverse those commits tomorrow and retest (I'm on
> development box).

I'm not convinced that Peter is barking up the right tree.  I'm noticing
that the profiles seem rather skewed towards parser/planner work; so I
suspect the contention is probably on access to system catalogs.  No
idea exactly why though.

It would be good to collect a few stack traces from the hung backends,
rather than relying on perf statistics.

			regards, tom lane

Commits

  1. Improve hash_create's API for selecting simple-binary-key hash functions.

  2. Fix WAL-logging of B-tree "unlink halfdead page" operation.

  3. Fix backup-block numbering in redo of b-tree split.

  4. Cleanup of new b-tree page deletion code.

  5. Fix bug in the new B-tree incomplete-split code.

  6. Fix bug in new B-tree page deletion code.

  7. Fix rm_desc routine of b-tree page delete records.

  8. Make the handling of interrupted B-tree page splits more robust.

  9. Fix race condition in B-tree page deletion.