Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-34843

Mongod can return operationTime greater than $clusterTime

    XMLWordPrintable

    Details

    • Type: Bug
    • Status: Closed
    • Priority: Major - P3
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 3.6.7, 4.0.0-rc0
    • Component/s: Sharding
    • Labels:
      None
    • Backwards Compatibility:
      Fully Compatible
    • Operating System:
      ALL
    • Backport Requested:
      v3.6
    • Sprint:
      Sharding 2018-05-21
    • Linked BF Score:
      15

      Description

      There is a race in the way mongod computes $clusterTime and operationTime. Before returning a response, mongod gets the latest cluster time from the LogicalClock and adds it to the request as $clusterTime (in appendReplyMetadata, called for successful commands here). Then, if a non null $clusterTime was computed, operationTime is computed by asking for the latest opTime on the client for writes, or the opTime of the last applied or committed write, for local and majority reads respectively. There is no synchronization that prevents the last applied or committed opTimes from advancing beyond the previously computed $clusterTime, allowing operationTime to be larger than $clusterTime in the response.

       A straightforward way to fix this could be to just compute operationTime before $clusterTime, because $clusterTime is always allowed to be greater than operationTime.

       

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              jack.mulrow Jack Mulrow
              Reporter:
              jack.mulrow Jack Mulrow
              Participants:
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved: