The Jenkins init script provided with the Debian package does not properly track the pidfile. It writes in jenkins.pid the pid of daemon instead of the spawned java process, that cause the restart/reload/stop targets to not work properly.
To make it worth, the init script provided for openSUSE/rpm and Debian are all different.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | hashar | T53817 /etc/init.d/jenkins script provided by Debian doesn't work properly | |||
| Resolved | hashar | T98976 Jenkins does not shutdown due to a deadlock in the IRC plugin |
Is that worth to upstream to http://bugs.debian.org ?
If so, which exact Jenkins version does this refer to?
$ cat /var/run/jenkins/jenkins.pid
32018
$
Pid 32018 is /usr/bin/daemon with --pidfile=/var/run/jenkins/jenkins.pid
Its children was pid 32020 which is the java process.
When using the init script 'stop' command, it send a signal to daemon and java process belong to jenkins:
killall -u jenkins java daemon
jenkins log:
Sep 24, 2013 7:46:23 AM winstone.Logger logInternal
INFO: JVM is terminating. Shutting down Winstone
Exception in thread "Channel reader thread: integration-debian-builder" java.lang.NullPointerException
at hudson.model.Computer.getNode(Computer.java:465)
at hudson.slaves.SlaveComputer.getNode(SlaveComputer.java:176)
at hudson.plugins.sshslaves.SSHLauncher.afterDisconnect(SSHLauncher.java:963)
at hudson.slaves.SlaveComputer$2.onClosed(SlaveComputer.java:419)
at hudson.remoting.Channel.terminate(Channel.java:749)
at hudson.remoting.SynchronousCommandTransport$ReaderThread.run(SynchronousCommandTransport.java:69)
But both daemon and java are still running after and the pidfile got deleted.
Have to kill -9 the process :(
We have moved to Debian Jessie and the init script is now managed by systemd. So I believe it is no more an issue and systemd would kill the java process for us.
For the record, we are now running Jenkins with systemd. So this definitely no more applies.