Uploaded image for project: 'Cassandra'
  1. Cassandra
  2. CASSANDRA-14722

New monotonic reads impl introduces risk during range movements

Agile BoardAttach filesAttach ScreenshotBulk Copy AttachmentsBulk Move AttachmentsVotersWatch issueWatchersCreate sub-taskConvert to sub-taskMoveLinkCloneLabelsUpdate Comment AuthorReplace String in CommentUpdate Comment VisibilityDelete Comments
    XMLWordPrintableJSON

Details

    • Correctness - Transient Incorrect Response
    • Critical

    Description

      Previously on digest mismatch we upgraded to CL.ALL, which harmed availability.  However, because we do not have consistent range movements, when they are underway it is possible for both the coordinator of a query and one of its targets have stale ring information. 

      In this case, we would previously have detected a digest mismatch and ensured ALL nodes responded before answering, so that any single stale node would have been detected.  In the new implementation, we try to get answers only from the nodes whose digests mismatched, so if the node were stale, it can respond with a stale data request also, reducing effective CL by one.

      Attachments

        Issue Links

        Activity

          This comment will be Viewable by All Users Viewable by All Users
          Cancel

          People

            Unassigned Unassigned Assign to me
            benedict Benedict Elliott Smith
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Slack

                Issue deployment