Uploaded image for project: 'Jenkins'
  1. Jenkins
  2. JENKINS-46499

Slaves sporadically not reverting to snapshot after limited builds count reached

    Details

    • Type: Bug
    • Status: Open (View Workflow)
    • Priority: Major
    • Resolution: Unresolved
    • Component/s: vsphere-cloud-plugin
    • Labels:
    • Environment:
      Debian GNU/Linux 8, AMD64
      Oracle JRE 1.8.0_144, AMD64
      Jenkins 2.60.3 (standalone/.deb installation)
      vSphere plugin 2.15
    • Similar Issues:

      Description

      Sometimes the vSphere-slaves do not shutdown+revert when configured to do so.

      This issue occurs sporadically, noticed this when being logged in on the slave using ssh and the slave did neither shutdown nor revert after the inspected build finished and the slave got listed as "offline".

      When selecting "reconnect" in Jenkins GUI, the slave gets reconnected without being reverted to the vm snapshot as desired.

      This issue occurrs sporadically, impacting different slaves on different vSphere clusters.

       

      After manually reverting the slave vm to its snapshot, then reconnecting the slave using Jenkins GUI, everything works fine again.

       

      when viewing the configuration page, the Jenkins logs show some NullPointerException, but this seems to be totally independent of the Exceptions occurring around the time the disconnect/revert fails to happen.

        Attachments

        1. Jenkins_VS_not_reverting.txt
          18 kB
          Tim-Christian Bloss
        2. Jenkins_VS_nullpointer.txt
          37 kB
          Tim-Christian Bloss

          Activity

          Hide
          alt_jmellor John Mellor added a comment - - edited

          I use limited job-count to force the node to revert and restart before every build, guaranteeing a clean build image.  This failure is causing build failures because of the unclean node.  In my site, it is happening around 50% of the time.  This should be a blocker bug.

          Also, not fixed in 2.16 version.

          Show
          alt_jmellor John Mellor added a comment - - edited I use limited job-count to force the node to revert and restart before every build, guaranteeing a clean build image.  This failure is causing build failures because of the unclean node.  In my site, it is happening around 50% of the time.  This should be a blocker bug. Also, not fixed in 2.16 version.

            People

            • Assignee:
              Unassigned
              Reporter:
              tcb_xy Tim-Christian Bloss
            • Votes:
              2 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

              • Created:
                Updated: