Обсуждение: Why are a lot of ROLLBACK queries in idle state

Hi,

When querying the pg_stat_acitivity I find a lot of ROLLBACK queries in idle
state. Can someone help me know why?

Thanks in advance.

select query, state, waiting, query_start from pg_stat_activity where query
ilike '%rollback%';
                                              query
| state  | waiting |          query_start

-------------------------------------------------------------------------------------------------+--------+---------+-------------------------------
 ROLLBACK
| idle   | f       | 2016-11-17 03:02:24.544345+00
 ROLLBACK
| idle   | f       | 2016-11-18 09:32:19.372843+00
 ROLLBACK
| idle   | f       | 2016-11-15 13:31:54.756704+00
 ROLLBACK
| idle   | f       | 2016-11-18 07:08:45.51883+00
 ROLLBACK
| idle   | f       | 2016-11-17 12:41:42.731161+00
 ROLLBACK
| idle   | f       | 2016-11-18 09:32:17.91838+00
 ROLLBACK
| idle   | f       | 2016-11-16 13:20:31.044959+00
 ROLLBACK
| idle   | f       | 2016-11-17 10:42:09.652842+00
 ROLLBACK
| idle   | f       | 2016-11-17 06:43:38.280102+00
 ROLLBACK
| idle   | f       | 2016-11-15 12:48:05.320185+00
 ROLLBACK
| idle   | f       | 2016-11-18 07:35:38.029441+00
 ROLLBACK
| idle   | f       | 2016-11-10 10:28:01.60524+00
 ROLLBACK
| idle   | f       | 2016-11-15 13:57:14.68677+00
 ROLLBACK
| idle   | f       | 2016-11-18 09:33:49.067333+00
 ROLLBACK
| idle   | f       | 2016-11-17 13:10:14.768031+00
 ROLLBACK
| idle   | f       | 2016-11-18 03:18:21.688177+00
 ROLLBACK
| idle   | f       | 2016-11-17 12:41:42.719949+00
 select query, state, waiting, query_start from pg_stat_activity where query
ilike '%rollback%'; | active | f       | 2016-11-18 09:55:05.476337+00
 ROLLBACK
| idle   | f       | 2016-11-18 08:24:59.286648+00
 ROLLBACK
| idle   | f       | 2016-11-17 13:10:14.767571+00
 ROLLBACK
| idle   | f       | 2016-11-18 09:32:19.448064+00
 ROLLBACK
| idle   | f       | 2016-11-15 13:57:14.680491+00
 ROLLBACK
| idle   | f       | 2016-11-11 09:18:26.057973+00
 ROLLBACK
| idle   | f       | 2016-11-15 12:24:52.661909+00
 ROLLBACK
| idle   | f       | 2016-11-11 09:18:26.064561+00
 ROLLBACK
| idle   | f       | 2016-11-18 06:56:57.2566+00
 ROLLBACK
| idle   | f       | 2016-11-11 09:16:21.458585+00
 ROLLBACK
| idle   | f       | 2016-11-16 13:21:02.534579+00




-----
--
Thanks,
Rajan.
--
View this message in context:
http://postgresql.nabble.com/Why-are-a-lot-of-ROLLBACK-queries-in-idle-state-tp5930917.html
Sent from the PostgreSQL - admin mailing list archive at Nabble.com.

rajan schrieb am 18.11.2016 um 10:55:
> When querying the pg_stat_acitivity I find a lot of ROLLBACK queries in idle
> state. Can someone help me know why?
>
> Thanks in advance.
>
> select query, state, waiting, query_start from pg_stat_activity where query
> ilike '%rollback%';

The _session_ is idle, not the query.

Quote from the manual:

If the session is idle, the "query" column of pg_stat_activity shows the last executed statement in that session.

So this simply means all those sessions properly ended their transaction using a ROLLBACK statement.

Thanks for the Reply.

If the _session_ is idle then shouldn't it be removed from pg_stat_activity?



-----
--
Thanks,
Rajan.
--
View this message in context:
http://postgresql.nabble.com/Why-are-a-lot-of-ROLLBACK-queries-in-idle-state-tp5930917p5930934.html
Sent from the PostgreSQL - admin mailing list archive at Nabble.com.

rajan schrieb am 18.11.2016 um 12:56:
> If the _session_ is idle then shouldn't it be removed from pg_stat_activity?

The session is only removed from pg_stat_activity when it's closed, i.e. when the client disconnects.

Being "idle" and disconnecting are two completely different things

thanks for the reply.

how do I kill these idle sessions? And if I do, will it cause any issues?



-----
--
Thanks,
Rajan.
--
View this message in context:
http://postgresql.nabble.com/Why-are-a-lot-of-ROLLBACK-queries-in-idle-state-tp5930917p5931144.html
Sent from the PostgreSQL - admin mailing list archive at Nabble.com.

rajan schrieb am 20.11.2016 um 07:38:
> thanks for the reply.
>
> how do I kill these idle sessions? And if I do, will it cause any issues?

Why do you think you need to kill them?

An idle connection is not really a problem and keeping them around for re-use is exactly what connection pools are
doing.

If you think those connections are a problem, you should rather investigate why your applications doesn't close, not
blindlykill them.