Fix FLCS cursor insert stalls or hangs for certain python tests

XMLWordPrintableJSON

    • Type: Bug
    • Resolution: Unresolved
    • Priority: Major - P3
    • None
    • Affects Version/s: None
    • Component/s: Column Store
    • None
    • Storage Engines, Storage Engines - Foundations
    • SE Foundations - Q3+ Backlog
    • 5

      In WT-13985, we disabled the FLCS portion of certain tests:

      • test_rollback_to_stable01.py
      • test_rollback_to_stable08.py
      • test_prepare_hs01.py

      and left a FIXME in place.  Preliminary analysis is in WT-13985.  In each of these cases, the test times out, and a message like:

      sweep-server: [WT_VERB_DEFAULT][WARNING]: Session 18 (@: 0x000001E11E1DB600 name: WT_CURSOR.cbt->dhandle) did not run a sweep for 60 minutes.

      appears on output.  This actually comes from the WT_CURSOR.insert call (WT-14971 describes how the name got mixed up).

      At any rate, the test ran for over 60 minutes, and after that during a sweep, we give an error like the above, showing the session was at that moment in cursor insert.  That doesn't mean necessarily that it's been sitting in insert for 60 minutes, but the fact that we've seen this many times in WT-13985 is a really strong indicator that it is in fact stuck in insert.

              Assignee:
              [DO NOT USE] Backlog - Storage Engines Team
              Reporter:
              Donald Anderson
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated: