Uploaded image for project: 'Jenkins'
  1. Jenkins
  2. JENKINS-46468

"no such computer" error occurs during SCM Pulling

      I am attempting to use a pipeline build triggered by github webhooks along with the kubernetes plugin.  I am seeing the following error in all of my poll logs:

      [poll] Latest remote head revision on refs/heads/master is: f4065637e80be6789d6d16c59ceee2cfde7b7ba9 - already built by 5

      ERROR: no such computer jenkins-slave-s10fr-lj69d

      Done. Took 0.58 sec No changes

      Looking at the plugin code there is a cache of what the workspace node has been in the past:

      https://github.com/jenkinsci/workflow-api-plugin/blob/63e8ad0c271573f4bebc57fb0776b3fac4fccea9/src/main/java/org/jenkinsci/plugins/workflow/FilePathUtils.java#L115

      Here is the line of code where the error is originating from:

      https://github.com/jenkinsci/workflow-job-plugin/blob/01f9e9bf32452a9d2ed7fa1e0ef92303e991a0c6/src/main/java/org/jenkinsci/plugins/workflow/job/WorkflowJob.java#L593

      Note that a workspace is required for the poll due to my use of "Additional Behaviors -> Polling ignores commits in certain paths"

      Assuming that a computer is available is an invalid assumption in the cloud where nodes are ephemeral.   I haven't yet come up with a work-around for this issue, so I am currently blocked.

      Thank you for your help!

          [JENKINS-46468] "no such computer" error occurs during SCM Pulling

          Sam Van Oort added a comment -

          abayer Executing discretion here... it feels like this ought to be do-able in some way with enough cleverness without checking out to a build agent, but it's definitely not a bug. So reclassifying and dropping priority because it's kind of edge-casey and potentially complex as well.

          Sam Van Oort added a comment - abayer Executing discretion here... it feels like this ought to be do-able in some way with enough cleverness without checking out to a build agent, but it's definitely not a bug. So reclassifying and dropping priority because it's kind of edge-casey and potentially complex as well.

          Andrew Bayer added a comment -

          svanoort - fwiw, if there's any possible resolution, it'd be to do what freestyle does (if I remember correctly) and kick off a build if polling requires a workspace but there isn't one available. Some polling options do always require a workspace, so that's what we've got to play with. shrug

          Andrew Bayer added a comment - svanoort - fwiw, if there's any possible resolution, it'd be to do what freestyle does (if I remember correctly) and kick off a build if polling requires a workspace but there isn't one available. Some polling options do always require a workspace, so that's what we've got to play with. shrug

          Gigi Jackson added a comment -

          We're running into this issue, although we're using the "polling ignores commits with certain message" additional behavior. I'm wondering if disabling lightweight checkout might be a work around. My understanding is that will cause the repo to be cloned on the master in order to get the Jenkinsfile. We use spot instances heavily for build agents, but master should always be up, and we'd definitely be willing to take the performance hit and do the comparison on a clone on master to get the build trigger behavior we want.

          Gigi Jackson added a comment - We're running into this issue, although we're using the "polling ignores commits with certain message" additional behavior. I'm wondering if disabling lightweight checkout might be a work around. My understanding is that will cause the repo to be cloned on the master in order to get the Jenkinsfile. We use spot instances heavily for build agents, but master should always be up, and we'd definitely be willing to take the performance hit and do the comparison on a clone on master to get the build trigger behavior we want.

          We have the same issue.

          Subscribing to this.

          Jan D'Hollander added a comment - We have the same issue. Subscribing to this.

          Turns out, I'm afflicted by the same problem as well. So I'm voting for some sort of resolution, even if it involves kicking off a build, no matter how short it is (in Kubernetes, spinning a VM up and down quickly is mostly acceptable).

          Zalan Meggyesi added a comment - Turns out, I'm afflicted by the same problem as well. So I'm voting for some sort of resolution, even if it involves kicking off a build, no matter how short it is (in Kubernetes, spinning a VM up and down quickly is mostly acceptable).

          allowing master nodes and putting this in the pipeline solved this for me:

              node('master') {
                  checkout scm
              }

          Balázs Hámorszky added a comment - allowing master nodes and putting this in the pipeline solved this for me: node('master') { checkout scm }

          Matt G added a comment -

          Same issue here. I'd almost say it's a serious bug because it can result in a build not happening, even when there are changes in SCM. It can be quite bad if the job is only configured to send notification to developers after a build failure. People may think they are covered by CI, but in reality it is not building any of their changes. (This is what happened to us)

          Matt G added a comment - Same issue here. I'd almost say it's a serious bug because it can result in a build not happening, even when there are changes in SCM. It can be quite bad if the job is only configured to send notification to developers after a build failure. People may think they are covered by CI, but in reality it is not building any of their changes. (This is what happened to us)

          Drew added a comment -

          Same issue here as well. This bug is a blocker for us as we have a larger number of packages that are subdirectories in a single git repository. I tried to use "Polling ignores commits in certain paths", but this leads to the same error as OP. Is the only workaround to use the master node for these sorts of builds?

          Drew added a comment - Same issue here as well. This bug is a blocker for us as we have a larger number of packages that are subdirectories in a single git repository. I tried to use "Polling ignores commits in certain paths", but this leads to the same error as OP. Is the only workaround to use the master node for these sorts of builds?

          Alan MD added a comment -

          Same issue here when using the EC2 dynamic agents. I opened another issue JENKINS-65199.
          Has anyone found a workaround?

          Alan MD added a comment - Same issue here when using the EC2 dynamic agents. I opened another issue  JENKINS-65199 . Has anyone found a workaround?

          Elijah Lynn added a comment - - edited

          Elijah Lynn added a comment - - edited Linking similar issues https://groups.google.com/g/jenkinsci-users/c/Ln5TsCYEr0A?pli=1 https://groups.google.com/g/jenkinsci-users/c/G9eQ9AeWf1k https://issues.jenkins.io/browse/JENKINS-65199

            Unassigned Unassigned
            wgorman Will Gorman
            Votes:
            13 Vote for this issue
            Watchers:
            21 Start watching this issue

              Created:
              Updated: