Multiwrite Replication with DISP Recovery (Multiwrite-DISP)

Multiwrite-DISP replication is a replication scheme that uses multiwrite replication for real-time updates and DISP for recovery.
cad1214
Multiwrite-DISP replication
is a replication scheme that uses multiwrite replication for real-time updates and DISP for recovery.
It is the recommended method for replication.
How Recovery Works in Multiwrite-DISP
When multiwrite replication with DISP recovery is enabled, multiwrite queues are not maintained for offline DSAs; any DSA that goes offline will rely on DISP recovery instead.
To enable multiwrite replication with DISP recovery use the following:
set multi-write-disp-recovery = true;
Note:
For the purposes of explanation, we call the DSA that is always available DSA 1, call the DSA that was unavailable and comes online, DSA 2.
When an unavailable DSA comes online, the following occurs:
  1. The multiwrite queue on DSA 1 for DSA 2 is enabled.
  2. DISP on DSA 1 performs the resynchronization by calculating and sending an update for the period in which DSA 2 was offline. While this is occurring, DSA 1 queues any further updates to DSA 2.
  3. When the DISP update completes, the multiwrite queue on DSA 1 is replayed to DSA 2.
  4. Multiwrite operation resumes.
Note:
There are no explicit DISP agreements; all resynchronization and reconciliation is automatic.
Benefits of DISP Recovery in Multiwrite-DISP
Enabling DISP recovery generally avoids manual resynchronization of the datastores.
DISP recovery works with multiwrite replication by handling all recovery operations. In effect, fast multiwrite is used during uptime and DISP is used only for recovery after a downtime.
DISP recovery has the following advantages:
  • Stable Calculation of Required Updates
    Updates are calculated from the original datastore, not queued in memory. This means that recovery can survive restarts of a master (although you can lose the in-memory queues).
  • State-based
    Resynchronization uses state-based updates. When there are a large number of replays, this is more efficient than replay-based recovery.
  • Reconciliation
    DISP processing supports automatic conflict resolution using a last-write-wins rule. Managing conflicts is often a problem for replay-based systems because the order of updates from different masters can be critical.
DISP recovery uses the highest authentication level that is available. Only specify the use of SSL authentication level (
ssl-auth
in a knowledge file) if you have set up the necessary SSL certificates for the DSAs in the region.
Enable Multiwrite-DISP
To turn on multiwrite with DISP recovery, use DXmanager, or add the following command to the DSA settings file:
set multi-write-disp-recovery = TRUE;
Granularity of DISP Reconciliation
When DISP resynchronization starts, it propagates all the changes to the DSA since the last successful multiwrite update.
In a multimaster system, it is possible for some changes to clash. In practice this is exceedingly rare. The window for discrepancy is small and all running master DSAs keep themselves synchronized using multiwrite. However, CA Directory resolves such conflicts automatically, using a last-write-wins rule. The granularity of reconciliation is an X.500 object.
Example: Conflict Resolution
If DSA 1 updates an object's surname attribute at time T, and DSA 2 updates an object's
phoneNumber
attribute at time T+1, then the final object after resynchronization is the object contained in DSA 2. This does not have the changes to the surname.