Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-5996

Review WT_SESSION_NO_LOGGING and if other flags should be retained when calling RTS

    • Type: Icon: Task Task
    • Resolution: Unresolved
    • Priority: Icon: Major - P3 Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • None

      Server restart hangs during recovery as part of rollback to stable checkpoint. More details are described in BF-16882 and WT-5988. The reason for wait forever is checkpoint is waiting for a logged records to be written to disk, but there is no log server exists during the recovery phase led to waiting forever.

      The issue was fixed by marking all the log records that are written during recovery unbuffered to let them write directly to disk instead of log server does.
      The issue was fixed by retaining the WT_SESSION_NO_LOGGING flag of the calling session when opening the rollback to stable internal session during recovery.

      Before adding rollback to stable code as part of DH changes, there exists a checkpoint that used to perform at the end of recovery and it is used to work fine without any problems.

      As part of this ticket, need to find out the difference between recovery checkpoint and rollback to stable checkpoint that is leading to wait.

            Assignee:
            backlog-server-storage-engines [DO NOT USE] Backlog - Storage Engines Team
            Reporter:
            haribabu.kommi@mongodb.com Haribabu Kommi
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

              Created:
              Updated: