AnsweredAssumed Answered

Disaster recovery from mirror volume without downtime

Question asked by communityadmin on May 21, 2014
Latest reply on Jun 3, 2014 by mbigelow
I'm setting up two M7 clusters in separate data centers via AWS EMR. The idea is that one cluster is the main "production" cluster, and the other "backup" cluster contains a remote mirror volume of the data from the production cluster.

Now, in the event of a disaster, say, the production cluster becomes unavailable, what I would like to do is this:

 1. Make the mirror volumes on the backup cluster read-write and have it act as the production cluster
 2. Once possible, rebuild the original production cluster and retrieve the data from the backup cluster (which is now acting as the production cluster)
 3. Once the restoration is performed on the new cluster, begin using it as the actual production cluster, and have the other cluster act as a mirror again.

The documentation seems to suggest that the only way to recover from the loss of a cluster or volume is to simply rebuild the cluster and copy the data from the mirror into the newly rebuilt cluster. My concern with this, however, is that there would be no way to write new data while the original cluster is being rebuilt, resulting in a lot of downtime (potentially).

Is it possible to switch a read-only mirror to read-write, and then back to read-only again? Or are there some other Best Practices for disaster recovery in the case where an entire cluster is lost?