ERROR: Cannot delete workspace: java.nio.file.FileSystemException #358
Comments
I used the Process Explorer to check if something is wrong with that file but it doesn't look like that. Then I rechecked with handle and voila. The only process which has a handle to that file is the running javaw.exe process. No other file has a handle open to it. That means that Java or Jenkins itself keeps a handle open to that file. Restarting Jenkins here was totally enough. I haven't had to delete the workspace at all. I assume that the same is also the case for all the other instances. Sadly I wasn't able to check the job which executed that testrun the last time on this box, given that it happened on Nov 2nd. I cannot access this anymore through the Jenkins CI. |
The Jenkins issue for this is https://issues.jenkins-ci.org/browse/JENKINS-19994 |
From looking at the plugin source code, it appears that we're hitting this exception in Jenkins core. I've added a log record for the plugin here: http://mm-ci-master.qa.scl3.mozilla.com:8080/log/Workspace%20Cleanup%20plugin/ so we should hopefully see a full stack trace next time this happens so we can find out where this exception is thrown. If we reset/restart production the logger may need to be re-added. This can be done by visiting http://mm-ci-master.qa.scl3.mozilla.com:8080/log/ and clicking 'Add new log recorder'. Name: 'Workspace Cleanup plugin', logger: 'hudson.plugins.ws_cleanup.PreBuildCleanup' |
Thanks Dave! If that problem doesn't reoccur until Monday, I will try to force it by restarting a win8 vm at home. I would assume it should be reproducible on other systems (beside production) too. |
This happened again out of sudden during the ondemand update tests today. The previous ondemand_update test from earlier today run fine without any errors or warnings. So I'm not sure what has been caused this. Restarting Jenkins slave on mm-win-8-32-1 fixed the problem again. |
Failure as seen in the Jenkins log for the workspace plugin: Nov 25, 2013 12:03:45 PM hudson.plugins.ws_cleanup.PreBuildCleanup preCheckout |
I have a hard time reading this stack. But if I see it correctly we already are failing before the cleanup_workspace plugin is called. Could this be a bug in Util.java? @davehunt what do you think? |
So in the above case the problematic file is again of size 0 and has been accessed the last time at 1:05am today. By that time we haven't run any job on that machine!! So some background task could have caused this problem. While thinking about that I can remember that we had a couple of rotate log failures. Not sure if that could be related. |
Indeed. The rotate log task is also failing: SEVERE: Failed to rotate log |
@davehunt would there be an easy way to check if one of those two try blocks are causing the 0 Byte size and leaving open handles behind? https://github.com/jenkinsci/jenkins/blob/1.509/core/src/main/java/hudson/Util.java#L275 |
I'm not sure, it's probably best to escalate this to the Jenkins dev team. |
Strangely I haven't found an issue for this. So filing it now.
It happened today again on Windows 8.1 64-1 when it tried to run functional tests for Aurora. It looks like this is the only testrun which is affected by this problem on this node. Other testruns were run fine, and even the aurora one succeeded on other nodes.
Here the output:
I have seen that the file in question has a size of 0.
The text was updated successfully, but these errors were encountered: