5.12. Resetting a Continuent Tungsten Dataservice

Follow these steps to reset replication for an entire dataservice. The current master will remain the master. Use the switch after completion to change the master.

See Section 5.10.6, “Resetting an Entire Dataservice from Filesystem Snapshots” if you would like to restore a file system snapshot to every server as part of this process.

  1. Put the dataservice into MAINTENANCE mode. This ensures that Continuent Tungsten will not attempt to automatically recover the service.

    cctrl> set policy maintenance
  2. Enable force mode:

    cctrl> set force true
  3. Shun each datasource:

    cctrl> datasource master shun
    cctrl> datasource slave1 shun
    cctrl> datasource slave2 shun
  4. Put each Tungsten Connector offline:

    cctrl> router * offline
  5. On each datasource, reset the service:

    shell> trepctl -service east offline
    shell> trepctl -service east reset -all -y
  6. Reconfigure the replicator and datasource configuration on each host, starting with the master:

    cctrl> set force true
    cctrl> replicator new-master master
    cctrl> replicator new-master online
    cctrl> datasource new-master master
    cctrl> datasource new-master online
    
    cctrl> replicator slave1 slave new-master
    cctrl> replicator slave1 online
    cctrl> datasource slave1 slave
    cctrl> datasource slave1 online
    
    cctrl> replicator slave2 slave new-master
    cctrl> replicator slave2 online
    cctrl> datasource slave2 slave
    cctrl> datasource slave2 online
  7. The connector can now be re-enabled and the cluster returned to operational state:

    cctrl> router * online
    cctrl> set policy automatic
    cctrl> cluster heartbeat

Any servers not matching the master must be reprovisioned. Use the tungsten_provision_slave tool to reprovision from the master or valid slave server.