Reasons for postgres processes beeing killed by SIGNAL 9?

Started by Clemens Eissereralmost 14 years ago5 messagesgeneral
Jump to latest
#1Clemens Eisserer
linuxhippy@gmail.com

Hi,

Recently single postgres processes are killed by SIGNAL 9 on our
virtual vvmware managed server without any manual interaction -
causing lost transactions.
Any ideas what could be the reason? Could postmaster the source of the signal?

We are running postgreql 8.4.7 on Linux 64-bit.

Thank you in advance, Clemens

2012-05-17 19:13:16 BST LOG: Serverprozess (PID 4849) wurde von
Signal 9 beendet: Killed
2012-05-17 19:13:16 BST LOG: aktive Serverprozesse werden abgebrochen
2012-05-17 19:13:16 BST khbldb_prod khbl WARNUNG: breche Verbindung
ab wegen Absturz eines anderen Serverprozesses
2012-05-17 19:13:16 BST khbldb_prod khbl DETAIL: Der Postmaster hat
diesen Serverprozess angewiesen, die aktuelle Transaktion
zurückzurollen und die Sitzung zu beenden, weil ein anderer
Serverprozess abnormal beendet wurde und möglicherweise das Shared
Memory verfälscht hat.
2012-05-17 19:13:16 BST LOG: alle Serverprozesse beendet; initialisiere neu
2012-05-17 19:13:16 BST LOG: Datenbanksystem wurde unterbrochen;
letzte bekannte Aktion am 2012-05-17 18:59:09 BST
2012-05-17 19:13:16 BST LOG: Datenbanksystem wurde nicht richtig
heruntergefahren; automatische Wiederherstellung läuft
2012-05-17 19:13:16 BST LOG: Redo beginnt bei 1/343D5010
2012-05-17 19:13:16 BST LOG: Datensatz mit Länge null bei 1/3445EAA8
2012-05-17 19:13:16 BST LOG: Redo fertig bei 1/3445EA78
2012-05-17 19:13:16 BST LOG: letzte vollständige Transaktion war
bei Logzeit 2012-05-17 19:13:13.575598+01

#2Steve Crawford
scrawford@pinpointresearch.com
In reply to: Clemens Eisserer (#1)
Re: Reasons for postgres processes beeing killed by SIGNAL 9?

On 05/17/2012 03:44 PM, Clemens Eisserer wrote:

Hi,

Recently single postgres processes are killed by SIGNAL 9 on our
virtual vvmware managed server without any manual interaction -
causing lost transactions.
Any ideas what could be the reason? Could postmaster the source of the signal?

We are running postgreql 8.4.7 on Linux 64-bit.

...

Out of memory or OOM killer?? Any such messages in system logs?

Cheers,
Steve

#3Clemens Eisserer
linuxhippy@gmail.com
In reply to: Steve Crawford (#2)
Re: Reasons for postgres processes beeing killed by SIGNAL 9?

Hi Steve,

Out of memory or OOM killer?? Any such messages in system logs?

That was my first thought too - but I could't find anything indicating
an OOM event in the logs.
Usually the server only uses ~110mb out of the available 2GB assigned to it.

So if this isn't a known postgres behaviour, I guess I have to dig a
bit in vmware related issues.

Thanks, Clemens

#4Clemens Eisserer
linuxhippy@gmail.com
In reply to: Clemens Eisserer (#3)
Re: Reasons for postgres processes beeing killed by SIGNAL 9?

Hi again,

We are still constantly getting postgresql processes killed by signal
9 from time to time, without any idea why or how.
Syslog seems completly clean.

In case a postgresql process would exceed some restricted resources
like file descriptors, would the kernel choose to terminate it using
SIGKILL? Are there any other common examples / occurences where
processes are terminated this way automatically?

Thank you in advance, Clemens

#5Alan Hodgson
ahodgson@simkin.ca
In reply to: Clemens Eisserer (#4)
Re: Reasons for postgres processes beeing killed by SIGNAL 9?

On Saturday, May 19, 2012 04:42:16 PM Clemens Eisserer wrote:

Hi again,

We are still constantly getting postgresql processes killed by signal
9 from time to time, without any idea why or how.
Syslog seems completly clean.

In case a postgresql process would exceed some restricted resources
like file descriptors, would the kernel choose to terminate it using
SIGKILL? Are there any other common examples / occurences where
processes are terminated this way automatically?

Check dmesg or the kernel log. I'd guess it's the OOM-killer. Assuming this is
on Linux, that is.