Uploaded image for project: 'Infrastructure'
  1. Infrastructure
  2. INFRA-1633

Stop building PR merges

    XMLWordPrintable

    Details

    • Type: Task
    • Status: In Progress (View Workflow)
    • Priority: Critical
    • Resolution: Unresolved
    • Component/s: ci
    • Labels:
      None
    • Similar Issues:

      Description

      I complained about this on Apr 26 but was told by R. Tyler Croy and Daniel Beck it was not really an issue. It is an issue. My plugin PR build has now been sitting for 47m waiting for an executor, the build queue has 111 items (https://ci.jenkins.io/load-statistics?type=min says it was up to >210 at one point), and all of our executors are rebuilding core PRs, one of which I see has not been touched since 2012. This is totally unreasonable. We really need to apply this setting at least to core builds, if not everywhere.

        Attachments

        1. 2019-03-23_22-37.png
          2019-03-23_22-37.png
          249 kB
        2. graph.png
          graph.png
          27 kB
        3. screenshot1.png
          screenshot1.png
          26 kB
        4. screenshot2.png
          screenshot2.png
          65 kB

          Issue Links

            Activity

            Hide
            danielbeck Daniel Beck added a comment -

            Don't know. Since I posted my comment a lot of things changed in core maintenance, I would not currently consider my preferences blocking.

            Show
            danielbeck Daniel Beck added a comment - Don't know. Since I posted my comment a lot of things changed in core maintenance, I would not currently consider my preferences blocking.
            Hide
            jglick Jesse Glick added a comment -

            Changing title to emphasize that core (jenkins) is not the only problem; as of this writing there are a bunch of dead ACI executors and all the working executors are tied up for the foreseeable future building acceptance-test-harness.

            Show
            jglick Jesse Glick added a comment - Changing title to emphasize that core ( jenkins ) is not the only problem; as of this writing there are a bunch of dead ACI executors and all the working executors are tied up for the foreseeable future building acceptance-test-harness .
            Hide
            jglick Jesse Glick added a comment -

            Mark Waite confirms that acceptance-test-harness is a problem in this regard.

            Show
            jglick Jesse Glick added a comment - Mark Waite confirms that acceptance-test-harness is a problem in this regard.
            Hide
            teilo James Nord added a comment - - edited

            this is probably worse for ATH than others - as the ATH job cancels previous builds for the current job when a new job starts.

            this causes a almost complete ATH run to be aborted which is a complete waste of resources for PRs especially if a PR itself has not been changed.

            Show
            teilo James Nord added a comment - - edited this is probably worse for ATH than others - as the ATH job cancels previous builds for the current job when a new job starts. this causes a almost complete ATH run to be aborted which is a complete waste of resources for PRs especially if a PR itself has not been changed.
            Hide
            teilo James Nord added a comment - - edited

            Another case in point is that this often results in PRs not having their incremental version deployed if something was merged to master and the PR is not at the head (which is the case if say you committed some PR feedback to change a typo).
            This has happened on a high proportion of my PRs to core (which is not statistically relevant)

            Show
            teilo James Nord added a comment - - edited Another case in point is that this often results in PRs not having their incremental version deployed if something was merged to master and the PR is not at the head (which is the case if say you committed some PR feedback to change a typo). This has happened on a high proportion of my PRs to core (which is not statistically relevant)

              People

              Assignee:
              oleg_nenashev Oleg Nenashev
              Reporter:
              jglick Jesse Glick
              Votes:
              2 Vote for this issue
              Watchers:
              8 Start watching this issue

                Dates

                Created:
                Updated: