Re: pg_stop_backup running for 10h?

[François Beausoleil <francois@xxxxxxxxxxx>]

Le 2014-01-12 à 05:38, Magnus Hagander a écrit :

On Sat, Jan 11, 2014 at 11:26 PM, François Beausoleil <francois@xxxxxxxxxxx> wrote:

Hi all,

I'm using OmniPITR to build a new slave. According to pg_stat_activity, pg_stop_backup has been running for nearly 11 hours. The WAL archive command is running just fine and reporting "Segment X successfully sent to all destinations".

I had the same issue almost a year ago (http://www.postgresql.org/message-id/9CC57302-10F8-4678-BBD3-028EC6B57051@xxxxxxxxxxx), but don't have permission issues this time around.

What could cause a pg_stop_backup() to run for such a long time?

Can't speak for the OmniPITR specific parts, but typically the archive_command reacting strangely would cause pg_stop_backup() to wait.

You include the logs from omniptr, but do you get anything in the *postgresql* logs? If it's the archive command it should clearly tell you that. It should also tell you if you can safely cancel the pg_stop_backup() command.

Oh well... pg_stop_backup() eventually finished by itself:

2014-01-12 05:57:06.174590 +0000 : 22722 : omnipitr-backup-master : LOG : Timer [SELECT pg_stop_backup()] took: 66049.684s

2014-01-12 05:57:06.296063 +0000 : 22722 : omnipitr-backup-master : LOG : pg_stop_backup('omnipitr') returned 2489/10004E78.

2014-01-12 05:57:06.409591 +0000 : 22722 : omnipitr-backup-master : LOG : Timer [Making data archive] took: 116525.818s

or 18 hours. That's long...

Postgres' logs themselves didn't have any mention of problems with the archive_command, so that's ruled out as well.

Thanks for taking the time to respond!

François

<<attachment: smime.p7s>>