Re: circular wait not triggering deadlock ? - Mailing list pgsql-general

From Justin Pryzby
Subject Re: circular wait not triggering deadlock ?
Date
Msg-id 20180308201301.GN24441@telsasoft.com
Whole thread Raw
In response to Re: circular wait not triggering deadlock ?  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
On Thu, Mar 08, 2018 at 03:05:36PM -0500, Tom Lane wrote:
> Justin Pryzby <pryzby@telsasoft.com> writes:
> > On Thu, Mar 08, 2018 at 01:57:06PM -0500, Tom Lane wrote:
> >> PID 20488 is evidently waiting for PID 6471 to finish its transaction.
> >> What's that one doing?
> 
> > Um, I thought I had kept track of all two pids but looks not..
> > query            | SELECT pg_export_snapshot();
> 
> pg_export_snapshot doesn't block for other transactions, though.
> Further down, you have output that confirms that:
> 
> >   6471 | idle in transaction | psql      | SELECT pg_export_snapshot();
> 
> That SELECT is the most recently completed command, not the current one.
> So now the question is what the connected application is waiting for
> before committing.

I believe it does an idle loop around sleep(1), until all the pg_dump
--snapshot children are done.

Justin


pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: circular wait not triggering deadlock ?
Next
From: Adrian Klaver
Date:
Subject: Re: What is the meaning of pg_restore output?