Bugzilla – Full Text Bug Listing
| Summary: | network issues on windows7tests | ||
|---|---|---|---|
| Product: | Community | Reporter: | Kim Moir <kim.moir> |
| Component: | CI-Jenkins | Assignee: | Eclipse Webmaster <webmaster> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | P3 | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Hardware: | PC | ||
| OS: | Windows XP | ||
| Whiteboard: | |||
| Bug Depends on: | |||
| Bug Blocks: | 295393 | ||
|
Description
Kim Moir
I'm restarting the slave. I was able to check out the basebuilder module fine on the command line(that sounds wrong but 'in the dos window' isn't any better). -M. I changed to build to just be the cvs checkout https://hudson.eclipse.org/hudson/view/Eclipse%20and%20Equinox/job/eclipse-JUnit/76/console It still has the same issue - can't check out the projects from cvs The build was still proceeding today after failling to checkout. I just tried to run another one and here is the error message I see Looks like the slave is still unhappy :-( FailedConsole Output View as plain text Started by user kmoir Building remotely on windows7tests hudson.util.IOException2: remote file operation failed: c:\Users\HUDSONBUILD\hudson\workspace\eclipse-JUnit at hudson.remoting.Channel@4de5a0cb:windows7tests at hudson.FilePath.act(FilePath.java:749) at hudson.FilePath.act(FilePath.java:735) at hudson.FilePath.mkdirs(FilePath.java:801) at hudson.model.AbstractProject.checkout(AbstractProject.java:1059) at hudson.model.AbstractBuild$AbstractRunner.checkout(AbstractBuild.java:479) at hudson.model.AbstractBuild$AbstractRunner.run(AbstractBuild.java:411) at hudson.model.Run.run(Run.java:1273) at hudson.model.FreeStyleBuild.run(FreeStyleBuild.java:46) at hudson.model.ResourceController.execute(ResourceController.java:88) at hudson.model.Executor.run(Executor.java:129) Caused by: java.net.SocketException: Broken pipe at java.net.SocketOutputStream.socketWrite0(Native Method) at java.net.SocketOutputStream.socketWrite(SocketOutputStream.java:92) at java.net.SocketOutputStream.write(SocketOutputStream.java:136) at java.io.BufferedOutputStream.flushBuffer(BufferedOutputStream.java:65) at java.io.BufferedOutputStream.write(BufferedOutputStream.java:109) at java.io.ObjectOutputStream$BlockDataOutputStream.drain(ObjectOutputStream.java:1838) at java.io.ObjectOutputStream$BlockDataOutputStream.writeByte(ObjectOutputStream.java:1876) at java.io.ObjectOutputStream.writeFatalException(ObjectOutputStream.java:1537) at java.io.ObjectOutputStream.writeObject(ObjectOutputStream.java:329) at hudson.remoting.Channel.send(Channel.java:419) at hudson.remoting.Request.call(Request.java:105) at hudson.remoting.Channel.call(Channel.java:557) at hudson.FilePath.act(FilePath.java:742) ... 9 more ERROR: Publisher hudson.tasks.Mailer aborted due to exception One build worked, now I'm getting timeouts again cvs [checkout aborted]: connect to dev.eclipse.org(172.25.25.51):2401 failed: Connection timed out FATAL: CVS failed. exit code=1 Hmm, that error was probably thrown as I was tweaking the DNS rules to make sure it was pointing to the right server. -M. Thanks, it seems be be working now I have a question. Running my build, if I echo %JAVA_HOME% it points to a unix install directory which seems weird because it's a windows install c:\Users\HUDSONBUILD\hudson\workspace\eclipse-JUnit>echo JAVA_HOME /shared/common/sun-jdk1.6.0_21_x64 JAVA_HOME /shared/common/sun-jdk1.6.0_21_x64 Another issue Could you copy all the vms on this machine to a directory that doesn't have a space in it like "Program Files" does. (https://bugs.eclipse.org/bugs/show_bug.cgi?id=296290#c30) The issue is that we run the p2.director like this to install the test bundles <java jar="${launcherPath}" failonerror="false" dir="${eclipse-home}" timeout="900000" fork="true" output="${basedir}/director.log" resultproperty="directorcode"> <arg line="-vm ${java.home}/bin/java" /> <arg line="-application org.eclipse.equinox.p2.director" /> <arg line="-consoleLog" /> <arg line="-flavor tooling" /> <arg line="-installIUs ${testPlugin},org.eclipse.test,org.eclipse.ant.optional.junit,org.eclipse.test.performance,org.eclipse.test.performance.win32,org.easymock" /> <arg line="-p2.os ${os}" /> <arg line="-p2.ws ${ws}" /> <arg line="-p2.arch ${arch}" /> <arg line="-roaming" /> <arg line="-profile SDKProfile" /> <arg line="-repository file:${repoLocation}" /> <arg line="-destination ${eclipse-home}" /> <arg line="-bundlepool ${eclipse-home}" /> </java> and it sees any space as the indicator of a new argument. I can't find an way to escape this that works. > I have a question. Running my build, if I echo %JAVA_HOME% it points to a unix > install directory which seems weird because it's a windows install That's the setting from the master. If you tell me which Java install should be 'JAVA_HOME' on the windows slave I'll set it's environment to match. (In reply to comment #7) > Could you copy all the vms on this machine to a directory that doesn't have a > space in it like "Program Files" does. Ok, I've copied them into c:\java (and changed the slave configure to match). -M. Thanks!
>>That's the setting from the master. If you tell me which Java install should
be 'JAVA_HOME' on the windows slave I'll set it's environment to match.
The jdk1.6.0_20 install please.
Can you also remove the "c:\program files" java location from the PATH environment variable and replace it with the "c:\java" one.
thanks, this is very helpful
Done. -M. Thanks for the changes! The network problems wrt to dev.eclipse.org have appeared again. The build just hangs... https://hudson.eclipse.org/hudson/job/eclipse-JUnit/97/console The windows slave seems to be offline again, could you please bring it back online. I've restarted the slave. Not sure what's the source of the network issues though. -M. I've successfully run our JUnit tests on the windows slave. Thanks for your help! One question: How much CPU and memory does the Windows 7 virtual instance have? If I run jobs on both Windows hudson instances, I assume we still have the same amount of resources and this won't make things faster. Is that correct? (The tests take about as long to complete as they do on our test box here so I'd like to make them faster). Right now it has 2 cpus and 2G of ram. More ram I can provide, but I don't have any cpus to spare. -M. Today at IBM we have two Windows JUnit machines. Each have 3GB of RAM. So if the eclipse.org Windows 7 install could have >= 6GB of RAM, we could see how much faster the tests run. I assume this requires bringing down the Hudson install so if you do this could you wait until my current job has finished since I'd like to check the test results. They have been running for four hours already :-) I've upped the ram to 8G and restarted things. -M. Seems to be having a network issue again https://hudson.eclipse.org/hudson/view/Eclipse%20and%20Equinox/job/eclipse-JUnit/116/console Windows slave seems to be having a network issue again - could you restart it? https://hudson.eclipse.org/hudson/view/Eclipse%20and%20Equinox/job/eclipse-JUnit/124/console I've restarted the slave. -M. This machine seems hung again. I can't see the workspace (just hangs) and a new job I started also just hangs and doesn't actually list any activity in the console output. Can you restart it? thanks :-) I've restarted the machine. -M. I just checked it and it says it's offline :-( OK that's irritating. Apparently if hudson thinks the slave has timed out but not 'disconnected' the auto-service startup fails. It's back now. -M. Is there a process that's holding on to the files? https://hudson.eclipse.org/hudson/view/Eclipse%20and%20Equinox/job/eclipse-JUnit/130/console It looks like the new build can't delete the old files in the the hudson workspace (In reply to comment #25) > Is there a process that's holding on to the files? I don't see any. I can stop the slave service clear the workspace and start it again if you'd like. -M. please do, thanks Matt. Done. -M. Has this been resolved? -M. Well, the windows slaves still seem to go offline every so often. Not sure what the root cause of this is..... Two things: The windows hudson instance is unresponsive again. Could you please restart it. Also, please rename the job eclipse-JUnit to eclipse-JUnit-Windows thanks :-) I didn't restart the host, just the hudson service. and it seems to be back now. I've renamed the job for you. -M. I think this can be closed. The windows machine has been very stable lately. Stay that way please :-) |