Replication failover/failback

[elangley]

Hello All,

I have a planned failover to a replica scheduled and I would like some input on the process, advantages/disadvantages and other thoughts. The failover will be temporary and failback employed.

System info:
Two FreeNAS servers both running 11.1-U7 in the same local area network.
System A replicates to System B
I want to use the System B replica to run production data.

Failover Options
It seems there are two options to do this. In both cases replication from the System A would be disabled prior to implementing.
Option A: clone the most recent replica snapshot and then mount the cloned snapshot for production use.
Option B: change the replicated dataset from read only to writable and mount the volume for production use.
Option C: ?

Failback Options - Copying the data back to System A from System is is not an option as it takes too long.
Option A: Create a replication job from Sytem B to System A.
Once the replication is completed it seems that the options are the same as when failing over, cloning or making the dataset writable. It seems that making the dataset writable would be cleaner because the other snapshots could be deleted and the dataset could be promoted.
Optin B: Not sure there is an Option B...

Let me know your thoughts and if you have done something like this.

~eric

 

[Heracles]

Hi elangley,

You do not need to disable the sync from A to B to do your test. Just keep it running. Because it is only a test, you do not need to revert the sync back from B to A.

Your option A would be to one to use:
You clone your latest snapshot on B and use it to complete your failover.
Once you confirmed your failover is working (read - write - delete data in the clone), you can stop the softwares and delete the clone.

And you are done.

If you wish to test the return from B to A, that is another case. So much that I actually recommend you not to do it and instead test a procedure as a failover from A to B followed by the re-deployment of a sync process from B to C. If you try to re-sync from B to A, you will probably end up deleting your own snapshots or data.

Good luck and congrats for deployind --and testing-- a complete failover solution,