-
Bug
-
Resolution: Fixed
-
Blocker
-
CentOS release 6.4 (Final)
OpenSSH_5.3p1, OpenSSL 1.0.0-fips 29 Mar 2010
Jenkins 1.532.1 LTS
apache-tomcat-6.0.37
ssh-slaves-plugin version 1.5
hi,
we experienced increasingly high amounts of CPU usage on our jenkins master machine:
i made a CPU sample using visualVM (screenshot attached)
the sample shows excessive CPU-time at CipherInputStream.fill_buffer() which performs the actual read from the ssh-connection as far as i understand this.
the sample only looks like this when a build is performed on a slave. I'm pretty sure the main cause of traffic via CipherInputStream is to stream logfiles of builds or other stuff that's getting archived.
the slaves are running on
CentOS release 6.4 (Final)
[kazesberger@grzcislave32 ~]$ ssh -V
OpenSSH_5.3p1, OpenSSL 1.0.0-fips 29 Mar 2010
the jenkins-master has no execution-slots itself.
we are using ssh-slaves-plugin version 1.5
currently i do not have any clue whether the problem is at trilead-ssh2-jenkins or at OpenSSH. the symptom occurs on the trilead-ssh2-jenkins side.
- is duplicated by
-
-
- Closed
-
- is related to
-
-
- Open
-
[JENKINS-22297] ssh-slave connection seems to have performance issues
Yves Martin
added a comment - After upgrade to 1.625.1, SSH slave communication is now as efficient as WebStart agent.
As a result, I consider this issue as fixed.
Yves Martin
added a comment - After upgrade to 1.625.1, SSH slave communication is now as efficient as WebStart agent.
As a result, I consider this issue as fixed.
I confirm I have the same problem.
I had to switch my Linux slaves to WebStart (JNLP) agent.
My master CPU consumption just drops to only 10% in average (and often less)
And as a consequence, my job executions are now three time faster.