Both new examples you provided seem consistent with getting a SIGTERM.
That means that all 3 examples you've shown happen within a second or
two of the start of a minute divisible by 5. In my mind that really does
add a lot of support to the suspicion of a cron job that is killing things.
> Feb 25, 2012 12:55:01 AM winstone.Logger logInternal
> INFO: JVM is terminating. Shutting down Winstone
> Feb 25, 2012 1:35:01 AM
> winstone.Logger logInternal
> INFO: JVM is terminating. Shutting down Winstone
Look in /etc/crontab, /etc/crontab.d/* and /var/spool/cron/crontabs for
any cron jobs that may be doing this.
You might also be able to find evidence of a cron job in /var/log/syslog
(or one of the other logfiles in /var/log depending on your clogging
config).
grep -i cron /var/log/syslog
grep 12:55: /var/log/syslog
grep 1:35: /var/log/syslog
Would be good start but if the logs have rotated in the meantime it may
be in one of the older syslog archive files.
Richard
On 25/02/2012 01:37, Domen Kožar wrote:
$ strace java -jar jenkins.war
...
INFO: Jenkins is fully up and running
) = -1 EINTR (Interrupted system call)
--- SIGTERM (Terminated) @ 0 (0) ---
futex(0xb79a1ae0, FUTEX_WAKE_PRIVATE, 1) = 1
rt_sigreturn(0xb716fbd8) = -1 EINTR (Interrupted system call)
futex(0xb716fbd8, FUTEX_WAIT, 20028, NULLFeb 25, 2012 1:35:01 AM
winstone.Logger logInternal
INFO: JVM is terminating. Shutting down Winstone
<unfinished ... exit status 143>
$
On Sat, Feb 25, 2012 at 12:11 AM, Sami Tikka <[email protected]
<mailto:[email protected]>> wrote:
Remove everything extra, like Jenkins startup scripts or extra
command line options. Download jenkins.war. Run: java -jar jenkins.war
Watch the output.
Another thing that comes to mind: strace java -jar jenkins.war
-- Sami
Domen Kožar <[email protected] <mailto:[email protected]>> kirjoitti 24.2.2012
kello 23.29:
With those settings, jvm still crashes. Maybe I should try with
Tomcat? I am still sure it's a memory thing, since I have 2gb of
ram on openvz and no swap. Any other suggestions?
On Fri, Feb 24, 2012 at 10:06 PM, Jeff <[email protected]
<mailto:[email protected]>> wrote:
My experience with Jenkins, Archiva and SONAR web applications
is that the default JVM settings regarding memory allocation
and PermGen sizes are not sufficient. I experienced weird
hanging/nonresponsive issues at various times where there were
no errors/exceptions, but things just stopped responding.
This was on Windows running Tomcat as a service.
My Windows VM's are configured with 4 GB memory and have set
the JVM settings as:
-Xmx2048m
-Xms512m
-XX:MaxPermSize=256m
-XX:+CMSClassUnloadingEnabled
-XX:+UseConcMarkSweepGC
The PermGen size was one of the bigger helps. I actually
caught an error in the logs that referenced an issue with
PermGen space when trying to shut down tomcat in one of my
attempts to find the cause. These settings have
performed well so far for all three web applications I'm
running. Maybe this will help.
On Fri, Feb 24, 2012 at 1:01 PM, Domen Kožar <[email protected]
<mailto:[email protected]>> wrote:
It is a bare debian 6 with jenkins installed. Is there any
reason jenkins would recieve signal 15? I guess it's time
to dig into Linux and find out a way where the signal is
coming from.
On Fri, Feb 24, 2012 at 7:59 PM, Mark Waite
<[email protected] <mailto:[email protected]>> wrote:
I'm a little surprised at that message. When I stop
the Jenkins process with "kill" (which defaults to
sending SIGTERM), I think it reports killing a process
named "java", not a process named "jenkins".
Is it possible that you have another program named
jenkins which is being started (and killed) instead of
"java -jar jenkins.war"?
Mark Waite
*From:* Richard Mortimer <[email protected]
<mailto:[email protected]>>
*To:* Domen Kožar <[email protected]
<mailto:[email protected]>>
*Cc:* [email protected]
<mailto:[email protected]>
*Sent:* Friday, February 24, 2012 11:26 AM
*Subject:* Re: Jenkins (JVM) crashes after few
minutes after start
On 24/02/2012 16:54, Domen Kožar wrote:
> That's the issue, java just exists, no
stacktrace, even if ran in
> non-daemon mode.
>
From you serverfault page it shows syslog output of
"Jan 31 17:50:02 cherry jenkins: jenkins: fatal:
client (pid 16189)
killed by signal 15, exiting"
That says that jenkins received signal 15 (SIGTERM
in most universes).
Now it is possible that jenkins sent itself that
but unlikely.
Can you check for cron jobs that might be sending
signals. If there is
nothing obvious you might want to look at the
timing of jenkins
terminating. That might give some pattern to allow
you to identify a
specific job killing things.
Richard
> On Fri, Feb 24, 2012 at 5:51 PM, Didier Durand
<[email protected]
<mailto:[email protected]>
> <mailto:[email protected]
<mailto:[email protected]>>> wrote:
>
> Hi,
>
> Do you have the Java stack trace of the crash?
>
> regards
>
> didier
>
> On Feb 24, 4:04 pm, Domen Kožar <[email protected]
<mailto:[email protected]> <mailto:[email protected]
<mailto:[email protected]>>>
> wrote:
> > I have been fighting with this one for a
month, would be really
> > pleased if someone can give tips or request
more data:
> >
> >
>
Crosspost:http://serverfault.com/questions/355516/jenkins-jvm-crashes-after-few...
>
>
--
Jeff Vincent
[email protected] <mailto:[email protected]>
See my LinkedIn profile at:
http://www.linkedin.com/in/rjeffreyvincent
I ♥ DropBox <http://db.tt/9O6LfBX> !!
