public inbox for [email protected]  

help / color / mirror / Atom feed

From: Benoit Lobréau <[email protected]>
To: Tom Lane <[email protected]>
Cc: 'Rajesh Kumar' <[email protected]>
Cc: [email protected]
Subject: Re: Commit with wait event on advisory lock!
Date: Thu, 27 Feb 2025 17:00:05 +0100
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>
References: <[email protected]>
	<CAJk5AtY-TWUYU7TR3KfbOyzvfDytTFTJfVyB1cKWpwKxNSV3Ww@mail.gmail.com>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>
	<[email protected]>

On 2/4/25 4:22 PM, Benoit Lobréau wrote:
> On 1/22/25 11:17 PM, Tom Lane wrote:
>>> By the way I also have commits which are waiting on ClientRead...
>>
>> That, on the other hand, is surely impossible.  I think maybe you
>> are misreading the stats display.  Typically I'd expect that such a
>> case indicates that the session is idle (awaiting a new command)
>> and the COMMIT is the last thing it did before that.
>>
>>             regards, tom lane
> 
> I can reproduce the issue using pgbench spamming "BEGIN; COMMIT;" and
> and running this query in psql:
> 
> SELECT DISTINCT state, wait_event, query
> FROM pg_stat_activity
> WHERE backend_type ILIKE '%client%'
> AND query ILIKE 'COMMIT%'
> \watch 0.5
> 
> After a short while I get the following :
> 
> active | ClientRead | COMMIT;
> 
> I looked into src/backend/utils/adt/pgstatfuncs.c and found that
> the state comes from the PgBackendStatus array, while the wait
> events are fetched from the proc array (using st_procpid taken
> from the backend status).
> 
> I don't think there is a guarantee that this "snapshot" is
> consistent across both arrays. It might just be a case of spamming
> pg_stat_activity and occasionally ending up with an "inconsistent
> snapshot."
> 
> Do you think this explanation holds weight?
> 
> I haven't been able to reproduce the advisory lock issue yet.
> 

Hi,

is this explanation for the active transaction with the ClientRead wait 
event plausible?

Thanks! I appreciate any insights you can share.

-- 
Benoit Lobréau
Consultant
http://dalibo.com

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: [email protected]
  Cc: [email protected], [email protected], [email protected], [email protected]
  Subject: Re: Commit with wait event on advisory lock!
  In-Reply-To: <[email protected]>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox