We recently updated our Jenkins master from 1.536 to 2.5 and we are now experiencing an intermittent recurring issue with PermGen errors that we did not have prior to the update. Occasionally one of our slaves (though not necessarily the same slave each time) will get the following error immediately after starting a build; 13:53:01 Building remotely on slavemach (linux) in workspace /home/Mike/workspace/MadiToNet 13:53:01 java.io.IOException: remote file operation failed: /home/Mike/workspace/MadiToNet at hudson.remoting.Channel@6e2f66:slavemach: java.io.IOException: Remote call on slavemach failed 13:53:01 at hudson.FilePath.act(FilePath.java:1014) 13:53:01 at hudson.FilePath.act(FilePath.java:996) 13:53:01 at hudson.scm.SubversionSCM.checkout(SubversionSCM.java:894) 13:53:01 at hudson.scm.SubversionSCM.checkout(SubversionSCM.java:830) 13:53:01 at hudson.scm.SCM.checkout(SCM.java:485) 13:53:01 at hudson.model.AbstractProject.checkout(AbstractProject.java:1269) 13:53:01 at hudson.model.AbstractBuild$AbstractBuildExecution.defaultCheckout(AbstractBuild.java:604) 13:53:01 at jenkins.scm.SCMCheckoutStrategy.checkout(SCMCheckoutStrategy.java:86) 13:53:01 at hudson.model.AbstractBuild$AbstractBuildExecution.run(AbstractBuild.java:529) 13:53:01 at hudson.model.Run.execute(Run.java:1741) 13:53:01 at hudson.model.FreeStyleBuild.run(FreeStyleBuild.java:43) 13:53:01 at hudson.model.ResourceController.execute(ResourceController.java:98) 13:53:01 at hudson.model.Executor.run(Executor.java:410) 13:53:01 Caused by: java.io.IOException: Remote call on slavemach failed 13:53:01 at hudson.remoting.Channel.call(Channel.java:789) 13:53:01 at hudson.FilePath.act(FilePath.java:1007) 13:53:01 ... 12 more 13:53:01 Caused by: java.lang.OutOfMemoryError: PermGen space ... and so on
A cmd->java -version reports that the slave is running jdk 1.8.0_91, which it seems doesn't support PermGen at all.
After restarting the slave machine (on a whim) our next build processed successfully until the very end, whereupon we recieved;
14:38:49 FATAL: java.io.IOException: Unexpected reader termination
14:38:50 FATAL: PermGen space
14:38:50 Finished: null
Interestingly, after restarting the master machine the issue has gone away for the time being. Is it the case that a PermGen error on the master can cause just one of several slaves to fail in this way? It seems that the master is using jdk 1.7.0_25, but as it isn't directly involved in the builds itself I had assumed that this would be ok?
Any help will be greatly appreciated,
Mike