Re: Postgres startup processes on linux? - resolved.

It's a normal pg connection server created by our own code. Nothing native to pg startup.

From: pgsql-general-owner(at)postgresql(dot)org [mailto:pgsql-general-owner(at)postgresql(dot)org] On Behalf Of Eshelman, James
Sent: Monday, March 30, 2009 3:29 PM
To: pgsql-general(at)postgresql(dot)org
Subject: [GENERAL] Postgres startup processes on linux?

We're running PG 8.1 on CentOS 5.0. When postgres starts the following processes apparently required by the DB itself get created:

postgres 23784 1 0 13:55 ? 00:00:00 /usr/bin/postmaster -p 5432 -D /var/lib/pgsql/data
postgres 23786 23784 0 13:55 ? 00:00:00 postgres: logger process
postgres 23788 23784 0 13:55 ? 00:00:00 postgres: writer process
postgres 23789 23784 0 13:55 ? 00:00:01 postgres: stats buffer process
postgres 23790 23789 0 13:55 ? 00:00:00 postgres: stats collector process
postgres 23802 23784 3 13:55 ? 00:00:58 postgres: airwave airwave [local] idle
postgres 23803 23784 0 13:55 ? 00:00:12 postgres: airwave airwave [local] idle

I understand what the first five are for. What are 23802 and 23803? These look like they might other processes started by the postmaster to serve connections. Tey have the property that they never go away, and seem to always follow *very closely* on the other pids, as though they were being created by the db itself, like the others just before them. However they the list information looks like a normal pg connection server created by postmaster. Our usual connection servers are identified in our own logs and do not include these two pids.

In addition pid 23802 exhibits the following mysterious behavior in the pg log, which is logging duration for all SQL statements running over 100 ms :

2009-03-30 14:58:23.763 EDT 25284 LOG: autovacuum: processing database "airwave"
2009-03-30 14:58:30.072 EDT 23802 LOG: duration: 236.061 ms statement: commit
2009-03-30 14:58:47.985 EDT 23802 LOG: duration: 269.877 ms statement: commit
2009-03-30 14:59:23.774 EDT 25305 LOG: autovacuum: processing database "airwave"
2009-03-30 14:59:33.923 EDT 23802 LOG: duration: 256.340 ms statement: commit
2009-03-30 14:59:51.130 EDT 23802 LOG: duration: 316.898 ms statement: commit
2009-03-30 15:00:07.967 EDT 23802 LOG: duration: 266.485 ms statement: commit
2009-03-30 15:00:23.790 EDT 25324 LOG: autovacuum: processing database "airwave"
2009-03-30 15:00:23.925 EDT 23802 LOG: duration: 338.512 ms statement: commit
2009-03-30 15:00:25.744 EDT 23802 LOG: duration: 272.170 ms statement: commit
2009-03-30 15:00:42.612 EDT 23802 LOG: duration: 310.099 ms statement: commit
2009-03-30 15:01:04.570 EDT 23802 LOG: duration: 254.574 ms statement: commit
2009-03-30 15:01:21.400 EDT 23802 LOG: duration: 245.433 ms statement: commit
2009-03-30 15:01:23.925 EDT 25345 LOG: autovacuum: processing database "airwave"
2009-03-30 15:01:38.286 EDT 23802 LOG: duration: 231.673 ms statement: commit
2009-03-30 15:02:12.017 EDT 23802 LOG: duration: 413.221 ms statement: commit
2009-03-30 15:02:23.954 EDT 25366 LOG: autovacuum: processing database "airwave"
2009-03-30 15:02:46.843 EDT 23802 LOG: duration: 268.504 ms statement: commit

You can see in the process list that process 23802 is building up a lot of cpu time, as confirmed in the pg log we run.

Can this be some native pg process opened by the DB itself, or is it some rogue connection server that we have to track down?

TIA,

Jim
HP ProCurve