Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-8923

Users with {foo:"1"} indexes able to rolling upgrade 2.2=>2.4, but then unable to initial sync

XMLWordPrintableJSON

    • Type: Icon: Improvement Improvement
    • Resolution: Done
    • Priority: Icon: Major - P3 Major - P3
    • None
    • Affects Version/s: 2.4.0-rc2
    • Component/s: Replication
    • Labels:
      None

      Assume a user has a 2.2 replica set with {foo:"1"} indexes on one or more of his/her collections. The user will successfully be able to perform a standard rolling upgrade to 2.4 and continue to use the indexes as ascending indexes. However, much later, if any of the members become stale and need a resync, the initial sync will fail:

      Fri Mar  8 21:13:56.669 [rsSync] replSet initial sync drop all databases
Fri Mar  8 21:13:56.669 [rsSync] dropAllDatabasesExceptLocal 1
Fri Mar  8 21:13:56.669 [rsSync] replSet initial sync clone all databases
Fri Mar  8 21:13:56.669 [rsSync] replSet initial sync cloning db: test
Fri Mar  8 21:13:56.671 [rsSync] replSet initial sync exception: 16736 Invalid index type '1' in index { v: 1, key: { foo: "1" }, ns: "test.foo", name: "foo_1" } 9 attempts remaining

      (One workaround: resync manually by performing a data file copy)

            Assignee:
            Unassigned Unassigned
            Reporter:
            rassi J Rassi
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: