Uploaded image for project: 'Evergreen'
  1. Evergreen
  2. EVG-14847

Allow separate test runners to output logs to different storage locations

    XMLWordPrintable

    Details

    • Type: New Feature
    • Status: Waiting (Blocked)
    • Priority: Major - P3
    • Resolution: Unresolved
    • Affects Version/s: None
    • Fix Version/s: next_quarter
    • Component/s: app
    • Labels:
      None

      Description

      Implementation notes from Brian Samek:

      At the time that this ticket was opened, Cedar supported a concept called group id. EVG-14608 will provide an implementation of logkeeper's job logs page using this group id. Resmoke sends logs to cedar with this group id.

      This ticket seeks to expand the group id so that other projects can also use the job logs page. The implementation would be to support a group id in the commands that upload test results. Evergreen code already has a place to pass the group id in options to cedar but currently does not. One possible interface would be a field in the commands that upload test results. Another would be some convention that if users name their test logs in a certain way, they would be uploaded with a group id based on the filename.

       

      Product notes from Maria van Keulen:

      Presently, the concept of group ID's is only supported for resmoke.py logs. These allow for the creation of the page described in EVG-14608 using Cedar.
      Per this comment, it'd be great if we could support a similar 'Job Logs' page for a sequentially-ordered list of test logs within a task. In the case where a task relies on separate runners for parallelization, these logs should ideally be separated per-runner. This is analogous to the way that resmoke.py separates logs based on job.
      It'd be great if a similar concept of group ID could be introduced to Evergreen test results commands so they can be grouped into sub-buckets within a task bucket in Cedar.

      The root ask here is to provide an interface to view test results in order of execution, and accommodate for the case where tests may be run by parallel runners. This is to help users debug issues where a test can cause side-effects for future tests. It's important to separate the tests by runner to isolate the source of the side effects.

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              backlog-server-evg Backlog - Evergreen Team
              Reporter:
              maria.vankeulen Maria van Keulen
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Dates

                Created:
                Updated: