Horton

Re: Latches vs lwlock contention

Heikki Linnakangas <hlinnaka@iki.fi>

View thread View on postgr.es
From: Heikki Linnakangas <hlinnaka@iki.fi>
To: Maxim Orlov <orlovmg@gmail.com>
Cc: Thomas Munro <thomas.munro@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>, Yura Sokolov <y.sokolov@postgrespro.ru>, Andres Freund <andres@anarazel.de>
Date: 2024-11-04T16:08:43Z
Lists: pgsql-hackers 
On 10/09/2024 19:53, Maxim Orlov wrote:
> I looked at the patch set and found it quite useful.
> 
> The first 7 patches are just refactoring and may be committed separately 
> if needed.
> There were minor problems: patch #5 don't want to apply clearly and the 
> #8 is complained
> about partitionLock is unused if we build without asserts. So, I add a 
> PG_USED_FOR_ASSERTS_ONLY
> to solve the last issue.
> 
> Again, overall patch looks good and seems useful to me. Here is the 
> rebased v5 version based on Heikki's patch set above.

Committed, thanks for the review!

In case you're wondering, I committed some of the smaller patches 
separately, but also squashed some of them with the main patch,  On 
closer look, the first patch, "Remove LOCK_PRINT() call that could point 
to garbage", wasn't fixing any existing issue. The LOCK_PRINT() was 
fine, because we held the partition lock. But it became necessary with 
the main patch, so I squashed it with that. And the others that I 
squashed were just not that interesting on their own.

The rest of Thomas's SetLatches work remains, so I left the commitfest 
entry in "Needs review" state.

-- 
Heikki Linnakangas
Neon (https://neon.tech)

Commits

  1. Fix crash if LockErrorCleanup() is called twice

  2. Split ProcSleep function into JoinWaitQueue and ProcSleep

  3. Move TRACE calls into WaitOnLock()

  4. Set MyProc->heldLocks in ProcSleep

  5. Fix comment in LockReleaseAll() on when locallock->nLock can be zero

  6. Allow a no-wait lock acquisition to succeed in more cases.

  7. Teach planner about more monotonic window functions