Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-50146

Removing a shard with 'uncommitted' documents in config.rangeDeletions on migration recipient can lead to incomplete state on donor

    XMLWordPrintable

    Details

    • Operating System:
      ALL
    • Backport Requested:
      v4.4

      Description

      The following scenario can occur:

      1. Shard X migrates a chunk to shard Y and completes
      2. At some point before the donor deletes the config.rangeDeletions document on the recipient, shard Y migrates that same chunk to some other shard and then gets removed
      3. Shard X receives ShardNotFound for either of these commands on the recipient and never updates its local config.rangeDeletions document. This will repeat even after failover, leading to permanent orphans and the inability to migrate an overlapping chunk back to shard X

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              backlog-server-sharding Backlog - Sharding Team
              Reporter:
              matthew.saltz Matthew Saltz
              Participants:
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

                Dates

                Created:
                Updated: