Uploaded image for project: 'Aurora'
  1. Aurora
  2. AURORA-1361

Explicit reconciliation must use SLAVE_ASSIGNED_STATES

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • None
    • 0.9.0
    • Scheduler
    • None
    • Twitter Aurora Q2'15 Sprint 5
    • 2

    Description

      TaskReconciler is currently using ACTIVE_STATES filter when fetching tasks for explicit reconciliation. Since Mesos does not know anything about non-assigned tasks, this results in log messages like this:

      allowed)
      E0619 00:09:35.219 THREAD175 org.apache.aurora.scheduler.state.TaskStateMachine$5.execute: Illegal state transition attempted: PENDING -> LOST (not allowed)
      I0619 00:09:35.220 THREAD13437 org.apache.aurora.scheduler.mesos.MesosSchedulerImpl.logStatusUpdate: Received status update for task 1371518881594-mesos-pending-resources-100-7
      7-fbbd6074-0896-495f-9a88-1d0ff48495fe in state TASK_LOST from SOURCE_MASTER with REASON_RECONCILIATION: Reconciliation: Task is unknown
      

      While there is no harm to PENDING or THROTTLED tasks, the above messages pollute logs and should be suppressed. Use SLAVE_ASSIGNED_STATES filter instead of ACTIVE_STATES.

      Attachments

        Activity

          People

            maximk Maxim Khutornenko
            maximk Maxim Khutornenko
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: