locked
Sync Knowledge appears to grow in spurts RRS feed

  • Question

  • I'm investigating an issue I have with n-tier synching.  The current code used WCF and SqlSyncProvider across about 20 replicas.  There is a central replica that communicates with every other replica and groups of replicas that can communicate with each other and the central pier.

    The issue is that occasionally for reasons I don't understand, the scope_sync_knowledge column grows and each replica effectively pulls down all the changes again.  For example, if this is the state of the replica keys to start

    0x0000000500000000000000040000000000000005000010
    00000004
    0764AB357C6C4CB3994978C1A93B10C7
    AFF0E8884C754611836C0E86E9589CD3
    8A7C1F60C03E400086EDA5494698ABEE
    9D7978800C914DFFAC6CC2EA1083700F
    000000180000100128020000010000001500000002000000010000000000000001
    00000004
    0000000000000000002F845E
    0000000100000000000AC0A0
    000000020000000000157302
    00000003000000000015730F
    ...

    After the event occurs, the scope_sync_knowledge will look like this:

    0x0000000500000000000000040000000000000005000010

    00000006
    0764AB357C6C4CB3994978C1A93B10C7
    AFF0E8884C754611836C0E86E9589CD3
    8A7C1F60C03E400086EDA5494698ABEE
    9D7978800C914DFFAC6CC2EA1083700F
    Some_new_guid><Some_new_guid>
    000000180000100128020000010000001500000002000000010000000000000001

    00000006

    0000000000000000002F845E
    0000000100000000000AC0A0
    000000020000000000157302
    00000003000000000015730F

    <new vector>

    <new vector>

    ...

    Also, once this occurs, each replica will start to pull changes (using batching) into the webbatching folder and reprocess all the changes again.  Is there some reason that the set of replicaids would get "lost" or recreated?  

    Thanks,

    Rob

    Wednesday, September 10, 2014 1:35 PM

Answers

  • Turns out I had several earlier versions of the client that were synching with this newer database and causing problems.  The new replicaIds were the older client replicas that were left connected to the network.
    • Marked as answer by rbeard47 Tuesday, September 16, 2014 1:30 AM
    Tuesday, September 16, 2014 1:30 AM

All replies

  • Turns out I had several earlier versions of the client that were synching with this newer database and causing problems.  The new replicaIds were the older client replicas that were left connected to the network.
    • Marked as answer by rbeard47 Tuesday, September 16, 2014 1:30 AM
    Tuesday, September 16, 2014 1:30 AM
  • the only time a new replica id would appear there is if a new replica is provisioned and synched. new scope = new replica id
    Wednesday, September 17, 2014 12:58 PM