Uploaded image for project: 'OpenDJ'
  1. OpenDJ
  2. OPENDJ-4935

Eliminate duplicate update messages in topologies containing 3 or more RSs


    • Type: Improvement
    • Status: Dev backlog
    • Priority: Major
    • Resolution: Unresolved
    • Affects Version/s: 6.0.0, 5.5.0, 4.0.0, 3.5.3
    • Fix Version/s: 7.0.0
    • Component/s: replication
    • Labels:


      When a topology contains three or more replication servers, there are a lot of messages of this kind appearing in the logs:

      category=SYNC severity=INFORMATION msg=Filtering out from log file [file] the record [message record] because it would break ordering. Last key appended is [key]

      For example, when running functional test:

       ./run-pybot.py -s replication_group3.failover DJ 

      that lasts approximately 15 minutes on my laptop,  there is more than one million of these messages occuring in the replication logs of the 3 servers.

      Cause of the messages

      This happens each time an update is performed on a DS, say DS1. The update is sent to RS1, which then send it to RS2 and RS3. Then, RS2 and RS3 will try to send the update to respectively to RS3 and RS2. Only the RS starting the first will actually send the update to the other.

      Total number of outdated messages sent in a topology

      The number of extra messages that could be avoided (messages that are send between all RS at the exclusion of the initial RS where changes come from) can be computed:

      N x (M-1) 

      for N messages and M RS in the topology (starting from M=3), because there is exactly N extra-messages exchanged beetwen any RS pair (excluding the initial RS when changes are coming).

      See OPENDJ-4823 for concrete examples with a topology of 3 RSs and 4 RSs.



          Issue Links



              • Assignee:
                nicolas.capponi@forgerock.com Nicolas Capponi
              • Votes:
                1 Vote for this issue
                5 Start watching this issue


                • Created: