Resynchronization best practices

The following best practices are recommended to improve performance and reduce the complexity and problems during resynchronization.

  • Use the QSYS2.SET_RESYNC_PRIORITIES procedure so that the objects most critical to your business operations are resynchronized sooner than lower priority objects. Any configuration of priorities applies to future resynchronization. See SET_RESYNC_PRIORITIES procedure for more information.
  • The user can specify a parallel degree to improve the performance of resynchronization if Db2 Symmetric Multiprocessing (SMP) is installed. However, since the primary node is still running applications, you should consider the performance impact to those applications. Parallel degree for resynchronization is set using the QSYS2.CHANGE_MIRROR procedure; see CHANGE_MIRROR procedure.
  • Do not vary off an IASP or suspend replication during resynchronization or long running DDL operations.
  • Do not hold any *EXCL, *EXCLRD, or *SHRNUP locks on objects on the secondary node during resynchronization. Such locks will prevent resynchronization operations from succeeding. The associated OTL entry will fail with a recoverable error, but the failure may also cause failures in dependent objects so that resynchronization will have to be performed again once the locks are released.
  • When resynchronization is running or while performing a warm clone, avoid long running database DDL operations, operations that would typically be handled by phase one processing, and RCL changes that replicate the same objects. These operations acquire locks that will cause resynchronization to run slower as it waits for locks to be released. Long running database DDL operations will also cause reprocessing of phase one entries and prevent resynchronization from progressing to phase two. If the locks are not released, related OTL entries can fail when resynchronization attempts to process the OTL entries.
  • If one of the mirrored nodes or IASPs has been unavailable for a very long amount of time, the OTL may have a large number of entries that need to be resynchronized. In that case, it may take less time to reclone the node or IASP than to perform resynchronization.