Обсуждение: Urgent! Log full with inexistent PID processes
Hi everybody.
We have a postgres 8.4.5-2 database under Debian Lenny 64bit.
This database has been working without any problem, but this afternoon server has got out of space due to a log dump.
Log file gets full with this entry until it reaches several gigabytes:
2011-06-27 13:25:07 CEST LOG: PID 21822 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21821 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21820 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21816 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21802 in cancel request did not match any process
Of course, we have restarted postmaster and rebooted machine but problem remains.
Any clue about this?
pg_hba only allows connection from localhost and, of course, there are no processes with these PIDs in the server...
Thanks in advance.
We have a postgres 8.4.5-2 database under Debian Lenny 64bit.
This database has been working without any problem, but this afternoon server has got out of space due to a log dump.
Log file gets full with this entry until it reaches several gigabytes:
2011-06-27 13:25:07 CEST LOG: PID 21822 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21821 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21820 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21816 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21802 in cancel request did not match any process
Of course, we have restarted postmaster and rebooted machine but problem remains.
Any clue about this?
pg_hba only allows connection from localhost and, of course, there are no processes with these PIDs in the server...
Thanks in advance.
Iñigo Martinez Lasala wrote:
Hi everybody.First check the space occupied by pg_log directory.
We have a postgres 8.4.5-2 database under Debian Lenny 64bit.
This database has been working without any problem, but this afternoon server has got out of space due to a log dump.
Log file gets full with this entry until it reaches several gigabytes:
2011-06-27 13:25:07 CEST LOG: PID 21822 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21821 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21820 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21816 in cancel request did not match any process
2011-06-27 13:25:07 CEST LOG: PID 21802 in cancel request did not match any process
Just delete the previous logs that are 1 month older.
rm -rf pg_log******
Of course, we have restarted postmaster and rebooted machine but problem remains.
Any clue about this?
pg_hba only allows connection from localhost and, of course, there are no processes with these PIDs in the server...
Thanks in advance.
* Iñigo Martinez Lasala: > Any clue about this? There is a misbehaving client which tries to issue broken cancel requests. You need to find it. > pg_hba only allows connection from localhost and, of course, there are > no processes with these PIDs in the server... Do you use TCP or UNIX domain sockets for the database connections? -- Florian Weimer <fweimer@bfk.de> BFK edv-consulting GmbH http://www.bfk.de/ Kriegsstraße 100 tel: +49-721-96201-1 D-76133 Karlsruhe fax: +49-721-96201-99
Hi Florian.
There is are no connections to DB. Only app using database is a web app that has been stopped also.
-----Original Message-----
From: Florian Weimer <fweimer@bfk.de>
To: Iñigo Martinez Lasala <imartinez@vectorsf.com>
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Urgent! Log full with inexistent PID processes
Date: Mon, 27 Jun 2011 13:40:49 +0000
There is are no connections to DB. Only app using database is a web app that has been stopped also.
-----Original Message-----
From: Florian Weimer <fweimer@bfk.de>
To: Iñigo Martinez Lasala <imartinez@vectorsf.com>
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Urgent! Log full with inexistent PID processes
Date: Mon, 27 Jun 2011 13:40:49 +0000
* Iñigo Martinez Lasala: > Any clue about this? There is a misbehaving client which tries to issue broken cancel requests. You need to find it. > pg_hba only allows connection from localhost and, of course, there are > no processes with these PIDs in the server... Do you use TCP or UNIX domain sockets for the database connections?
* Iñigo Martinez Lasala: > There is are no connections to DB. Only app using database is a web app > that has been stopped also. Uh-oh. Can you "strace -f" the postmaster? Surely something is connecting to it. -- Florian Weimer <fweimer@bfk.de> BFK edv-consulting GmbH http://www.bfk.de/ Kriegsstraße 100 tel: +49-721-96201-1 D-76133 Karlsruhe fax: +49-721-96201-99
You are my man, bud. :)
A remote server was launching a process against this database. We had limited access via pg_hba.conf (or at least we thought access was limited). Since a database admin was modifiying config in paralell with me, we have a mess with config and so access restriction was not applying.
Thank you very mucho.
-----Original Message-----
From: Florian Weimer <fweimer@bfk.de>
To: Iñigo Martinez Lasala <imartinez@vectorsf.com>
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Urgent! Log full with inexistent PID processes
Date: Tue, 28 Jun 2011 09:25:26 +0000
A remote server was launching a process against this database. We had limited access via pg_hba.conf (or at least we thought access was limited). Since a database admin was modifiying config in paralell with me, we have a mess with config and so access restriction was not applying.
Thank you very mucho.
-----Original Message-----
From: Florian Weimer <fweimer@bfk.de>
To: Iñigo Martinez Lasala <imartinez@vectorsf.com>
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Urgent! Log full with inexistent PID processes
Date: Tue, 28 Jun 2011 09:25:26 +0000
* Iñigo Martinez Lasala: > There is are no connections to DB. Only app using database is a web app > that has been stopped also. Uh-oh. Can you "strace -f" the postmaster? Surely something is connecting to it.