Core Server
  1. Core Server
  2. SERVER-4373

Windows service should "do the right thing" after a crash

    Details

    • Type: Bug Bug
    • Status: Open Open
    • Priority: Major - P3 Major - P3
    • Resolution: Unresolved
    • Affects Version/s: None
    • Fix Version/s: Planning Bucket B
    • Component/s: Admin
    • Environment:
      Windows, mongod.exe installed as a Windows service using "mongod --install"
    • Backport:
      No
    • Operating System:
      Windows
    • Bug Type:
      Logical
    • # Replies:
      2
    • Last comment by Customer:
      false

      Description

      When MongoDB is installed as a service on Windows, users will have an expectation that the database will keep itself available across system restarts. It does this fine when the system is shut down in an orderly way, but on a power failure or bluescreen or system crash, mongod.exe can't start up without some handholding because of the leftover mongod.lock file. If we just auto-restart, we fail immediately because of the leftover file. If we don't auto-restart, then users will wonder why we're not running and have to go do something to fix it.

      It would be a lot better if we had a scheme for climbing out of this ditch and could make the code just recover automatically (when possible) and issue notifications to administrators to please come and intervene when that is required.

        Issue Links

          Activity

          Hide
          Tad Marshall
          added a comment -

          This bug is inspired by (fallout from?) SERVER-3582 – Mongo Windows Service locks up CPU when computer reboots after a crash. In fixing that bug, I changed our behavior from "infinite failure loop" to "do nothing". It would be much better to "do the right thing", once we figure out what that is.

          Show
          Tad Marshall
          added a comment - This bug is inspired by (fallout from?) SERVER-3582 – Mongo Windows Service locks up CPU when computer reboots after a crash. In fixing that bug, I changed our behavior from "infinite failure loop" to "do nothing". It would be much better to "do the right thing", once we figure out what that is.
          Hide
          Eliot Horowitz
          added a comment -

          tricky to get right

          Show
          Eliot Horowitz
          added a comment - tricky to get right

            People

            • Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

              • Created:
                Updated:
                Days since reply:
                2 years, 20 weeks, 5 days ago
                Date of 1st Reply: