Host Failure

In this scenario, the host on which the application is running is lost and eventually all hosts in the system zone or cluster are lost.

  To perform the host failure test

1. Shut down the host on which the application is running:

   The service group fails over to hostb and devices are in the P-VOL/PAIR state.

2. Halt or shut down hostb.

   In a replicated data cluster, the group fails over to hostc or hostd depending on the value of the FailOverPolicy attribute in the cluster.

   In a global cluster, a cluster down alert appears and gives you the opportunity to fail over the service group manually.

   In both environments, the devices on the target array remain S-VOLs because they cannot communicate with the original primary's RAID manager, but they transition to the writable SSWS status. Also, the failover may take some time as the RAID manager connection times out.

3. Reboot the two hosts that were shut down. A swap resynchronization is required to demote the original P-VOLs:
   # hares -action HTCRes pairresync-swaps -sys system
4. Migrate back when the devices transition from COPY to PAIR:
   1. In the Service Groups tab of the Cluster Explorer configuration tree, right-click the service group.
   2. Click Switch To, and click the system where the service group was initially online (hosta).

   The service group comes online on hosta and devices swap roles again.