Re: Lock Wait Statistics (next commitfest)

From: Mark Kirkwood <markir(at)paradise(dot)net(dot)nz>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: Jaime Casanova <jcasanov(at)systemguards(dot)com(dot)ec>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Lock Wait Statistics (next commitfest)
Date: 2009-07-25 00:22:50
Message-ID: 4A6A505A.5070504@paradise.net.nz
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Tom Lane wrote:
> Mark Kirkwood <markir(at)paradise(dot)net(dot)nz> writes:
>
>> Yeah, enabling log_lock_waits is certainly another approach, however you
>> currently miss out on those that are < deadlock_timeout - and
>> potentially they could be the source of your problem (i.e millions of
>> waits all < deadlock_timeout but taken together rather significant).
>> This shortcoming could be overcome by making the cutoff wait time
>> decoupled from deadlock_timeout (e.g a new parameter
>> log_min_lock_wait_time or similar).
>>
>
> The reason that they're tied together is to keep from creating
> unreasonable complexity (and an unreasonable number of extra kernel
> calls) in management of the timeout timers. You will find that you
> can't just wave your hand and decree that they are now decoupled.
>
>

Thanks Tom - I did wonder if there was a deeper reason they were tied
together!

Cheers

Mark

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Dave Page 2009-07-25 00:24:28 Re: Proposal: More portable way to support 64bit platforms
Previous Message KaiGai Kohei 2009-07-25 00:16:47 Re: SE-PostgreSQL Specifications