A newer version of this documentation is available.

View Latest

Enabling Timestamp-based Conflict Resolution for Migrated Data

      +
      The Timestamp-based Conflict Resolution is a new conflict resolution type added in version 4.6.0. This new feature is supported for new buckets that are created in Couchbase Server version 4.6.0. You cannot change the conflict resolution mode to the Timestamp-based Conflict Resolution for existing buckets after upgrading to version 4.6.0.

      If you wish to enable the timestamp-based conflict resolution for your existing data, then you must migrate your data to version 4.6.0 cluster using the cbbackupmgr tool. To learn more about the tool, See Backup and Restore.

      This is a one time migration and the bucket must be switched to a new conflict resolution type as part of the migration.

      To understand more about timestamp-based Conflict Resolution, see Timestamp-Based Conflict Resolution.

      Here are two scenarios:

      • Scenario 1 - In case of unidirectional replication, for example, for a disaster recovery where Cluster 1 (Bucket A) has one way replication streams to Cluster 2 (Bucket A’).

      • Scenario 2 - In case of bidirectional replication, for example, for data locality where Cluster 1 (Bucket A) has both ways replication streams to Cluster 2 (Bucket A’).

      For Scenario 1, perform the following steps to migrate your data to a new cluster:

      1. Stop application traffic coming into Cluster 1 (Bucket A).

        Allow enough time for the replication queues to drain. Check the outbound XDCR mutation statistics to confirm. For instructions, see Monitoring Outgoing XDCR.
      2. Stop and delete Replication stream to Cluster 2 (Bucket A’). For instructions, see XDCR Management Overview.

      3. Run the cbbackupmgr tool to back up the entire bucket’s (Bucket A) data. For instructions, see Backup.

      4. Delete Bucket A from Cluster 1. For instructions, see Delete a Bucket.

      5. Upgrade Cluster 1 to Couchbase Server version 4.6.0. For instructions, see Upgrading Couchbase Server.

      6. Create a new Bucket B with the conflict resolution type as Timestamp selected. For instructions, see Create a Bucket.

      7. Run the cbbackupmgr tool to restore data. When restoring data from backup (use the --force-updates option). Make sure to disable the Conflict Resolution option during the restore. This is required since the conflict resolution types of the source and destination clusters do not match.

      8. Once the restore operation is completed on Cluster 1, delete Bucket A’ from Cluster 2.

      9. Upgrade Cluster 2 to Couchbase Server version 4.6.0.

      10. Create a new Bucket B’ with the Conflict Resolution type as Timestamp selected.

      11. Create a new unidirectional replication for Bucket B. See XDCR Management Overview.

      For Scenario 2, perform the following steps to migrate your data to a new cluster:

      1. Stop application traffic coming to Cluster 1 (Bucket A) and Cluster 2 (Bucket A’).

        Note: Allow enough time for the replication queues to drain. Check the outbound XDCR mutation statistics to confirm. For instructions, see Monitoring Outgoing XDCR.
      2. Stop and delete replication to both clusters. For instructions, see XDCR Management Overview.

      3. Run the cbbackupmgr tool to backup the entire bucket data on both clusters. For instructions, see Backup.

      4. Delete the Bucket A from Cluster 1 and Bucket A’ from Cluster 2. For instructions, see Delete a Bucket.

      5. Upgrade both clusters to Couchbase Server version 4.6.0. For instructions, see Upgrading Couchbase Server.

      6. Create new buckets on both clusters with the conflict resolution type as Timestamp selected. For instructions, see Create a Bucket.

      7. Run the cbbackupmgr tool to restore data. When restoring data from backup (use the --force-updates option). Make sure to disable Conflict Resolution option during the restore. This is required because the conflict resolution types of the source and destination do not match.

      8. Once the restore operation is completed on both clusters, create replication streams both ways from Cluster 1 and Cluster 2. See XDCR Management Overview.