Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-10755

timestamp_abort should do more work after recovering a database

    • Type: Icon: Improvement Improvement
    • Resolution: Fixed
    • Priority: Icon: Major - P3 Major - P3
    • WT11.2.0, 7.0.0-rc0
    • Affects Version/s: None
    • Component/s: None

      Summary

      timestamp_abort (and our other _abort tests) doesn't do any more work on the database after it recovers it and verifies it after crash. The test should do more work on the database in order to uncover any potential issues that might happen only after recovering from an unclean shutdown, which, for example, seems to be the case of HELP-41498.

      Acceptance criteria

      The test program should take an additional command-line argument that specifies the number of iterations that the test should run, where an iteration means running a workload, crashing, and recovering. For example, running three iterations means that we would crash the database three times, and we would run a workload on a formerly crashed database twice.

      Possible approach

      The work/crash/recovery/verify part of the main() function should be in a for loop. For simplicity, we should modify the initial keys and timestamps so that there is no overlap between iterations. (Having overlap between iterations might be an interesting extension later.)

            Assignee:
            peter.macko@mongodb.com Peter Macko
            Reporter:
            peter.macko@mongodb.com Peter Macko
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: