Uploaded image for project: 'Jenkins'
  1. Jenkins
  2. JENKINS-43607

Jenkins pipeline not aborted when the machine running docker container goes offline

    Details

    • Type: New Feature
    • Status: Resolved (View Workflow)
    • Priority: Major
    • Resolution: Duplicate
    • Labels:
      None
    • Environment:
      Jenkins ver. 2.53
      Pipeline job /
      Pipeline: Nodes and Processes plugins : ver. 2.10
    • Similar Issues:

      Description

       Preconditions

      Jenkins pipeline job is configured to run parallel actions in different docker swarm nodes.

      Procedure

      1. Run job
      2. Force disconnect of a node running a part of this job

      Actual outcome

      Job will never terminate. The pipeline part will remain stuck in:

      Cannot contact swarm-xxxxxxxx: hudson.remoting.RequestAbortedException: java.nio.channels.ClosedChannelException

      The exception is catched by workflow-durable-task-step-plugin and used to display the log above.

      Expected outcome

      The pipeline part execution should generate an exception that can be catched.

      This will can allow implementing a retry strategy in Pipeline job.

        Attachments

          Issue Links

            Activity

            Hide
            jglick Jesse Glick added a comment -

            As with JENKINS-36013, currently the model is that a node may go offline and later be reconnected, in which case the step will quietly resume printing output and exit normally. For Swarm or other cloud-like node schemes, a disconnection may be followed by an actual permanent removal of the node definition, in which case it would be desirable for the step to abort.

            Show
            jglick Jesse Glick added a comment - As with  JENKINS-36013 , currently the model is that a node may go offline and later be reconnected, in which case the step will quietly resume printing output and exit normally. For Swarm or other cloud-like node schemes, a disconnection may be followed by an actual permanent removal of the node definition, in which case it would be desirable for the step to abort.
            Hide
            stickycode Michael McCallum added a comment -

            Jesse Glick  should this get more attention? there are a number of tickets and questions turning up online as  ephemeral nodes are becoming way more common. GKE in particular makes its very cheap and easy.

            Show
            stickycode Michael McCallum added a comment - Jesse Glick   should this get more attention? there are a number of tickets and questions turning up online as  ephemeral nodes are becoming way more common. GKE in particular makes its very cheap and easy.

              People

              • Assignee:
                Unassigned
                Reporter:
                aymen_parrot Aymen Bouaziz
              • Votes:
                1 Vote for this issue
                Watchers:
                5 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: