From: | Heikki Linnakangas <heikki(dot)linnakangas(at)enterprisedb(dot)com> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | Robert Haas <robertmhaas(at)gmail(dot)com>, pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: LWLOCK_STATS |
Date: | 2012-01-07 21:48:23 |
Message-ID: | 4F08BDA7.4090203@enterprisedb.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On 07.01.2012 19:18, Tom Lane wrote:
> Heikki Linnakangas<heikki(dot)linnakangas(at)enterprisedb(dot)com> writes:
>> A couple of weeks ago I wrote a little patch that's similar to
>> LWLOCK_STATS, but it prints out % of wallclock time that is spent
>> acquiring, releasing, or waiting for a lock. I find that more useful
>> than the counters.
>
> I would think that the measurement overhead required to obtain two
> wall-clock values for every LWLock touch would be so high as to render
> any results from this quite suspect.
It's based on sampling. The timer calls a callback every X ms, which
checks if it's waiting for any lock at that moment, and bumps a counter
if so. In LWLockAcquire/Release you just set/reset a global status variable.
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com
From | Date | Subject | |
---|---|---|---|
Next Message | Peter Geoghegan | 2012-01-08 06:10:06 | Re: Progress on fast path sorting, btree index creation time |
Previous Message | Tom Lane | 2012-01-07 21:41:42 | Re: return values of backend sub-main functions |