From: | Andres Freund <andres(at)2ndquadrant(dot)com> |
---|---|
To: | Heikki Linnakangas <hlinnakangas(at)vmware(dot)com> |
Cc: | Merlin Moncure <mmoncure(at)gmail(dot)com>, Ants Aasma <ants(at)cybertec(dot)at>, Peter Eisentraut <peter_e(at)gmx(dot)net>, Дмитрий Дегтярёв <degtyaryov(at)gmail(dot)com>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: [PERFORM] Cpu usage 100% on slave. s_lock problem. |
Date: | 2013-11-21 15:02:00 |
Message-ID: | 20131121150200.GC31748@awork2.anarazel.de |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On 2013-11-21 16:25:02 +0200, Heikki Linnakangas wrote:
> Hmm. All callers of RecoveryInProgress() must be prepared to handle the case
> that RecoveryInProgress() returns true, but the system is no longer in
> recovery. No matter what locking we do in RecoveryInProgress(), the startup
> process might finish recovery just after RecoveryInProgress() has returned.
True.
> What about the attached? It reads the shared variable without a lock or
> barrier. If it returns 'true', but the system in fact just exited recovery,
> that's OK. As explained above, all the callers must tolerate that anyway.
> But if it returns 'false', then it performs a full memory barrier, which
> should ensure that it sees any other shared variables as it is after the
> startup process cleared SharedRecoveryInProgress (notably,
> XLogCtl->ThisTimeLineID).
I'd argue that we should also remove the spinlock in StartupXLOG and
replace it with a write barrier. Obviously not for performance reasons,
but because somebody might add more code to run under that spinlock.
Looks good otherwise, although a read memory barrier ought to suffice.
Greetings,
Andres Freund
--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
From | Date | Subject | |
---|---|---|---|
Next Message | Bruce Momjian | 2013-11-21 15:05:15 | Status of FDW pushdowns |
Previous Message | Dean Rasheed | 2013-11-21 14:55:13 | Re: WIP patch for updatable security barrier views |