Uploaded image for project: 'HPCC'
  1. HPCC
  2. HPCC-10917

A broken MP connection can go unnoticed and deadlock

    Details

    • Type: Bug
    • Status: Resolved
    • Priority: Not specified
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 4.2.4
    • Component/s: Thor
    • Labels:
      None

      Description

      A broken socket is detected by the MP layer and passed onto receivers pending on messages.
      If a receiver is waiting on a specific endpoint, it will get notified of the close and abort, ultimately causing a 'MP link closed' error to be propagated through.

      However, if a receiver is waiting on a RANK_ALL, then the close notification mechanism fails to pass on the fact that one of the endpoints the receiver is interested in has closed.

      This fix will pass on the closed endpoint to the receiver of a RANK_ALL.

        Attachments

          Activity

            People

            • Assignee:
              jakesmith Jake Smith
              Reporter:
              jakesmith Jake Smith
            • Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: