Why does Failover default to leaving Source hosts running?

[garymansell]

I am struggling to understand why the default option for failover and "failover plans" is to bring up the replica servers without shutting down the source hosts - please can someone help me better my understanding of this?

It seems to me that one would never want to do this, as the replica and the source machines would divert in state from each other immediately they are started. Further, even if this was just for testing that the replica was actually boot-able, then the very action of bringing it up by means of the failover will either cause an IP clash and hence end users could be accessing both the source and replica machines concurrently, or if a Re-IP rule is configured for the replicated VM then a DNS issue will occur where the Re-IP address will overwrite the source address in DNS and get replicated around the Domain, causing end users to no longer be able to access the source host. I would also be a bit concerned with two identical machines being active on the Domain at the same time that might possibly cause corruption in AD.

It seems to me that one would only ever want to be able to do a "planned failover" that stops and fails over the source host (with latest changes or not) followed by either an "undo failover" (to revert back to source losing replica changes) or a "planned failback" (to bring replica changes back to source).

What am I missing here? Why is this even an option, let alone the default one?

Thanks in advance

Gary

[garymansell]

Further to this - how do you do a "Planned Failover" whereby the source machines are replicated and shutdown, before the replica's are started when using a Failover Plan (as the vm's involved depend on each other)?

[garymansell]

I have realised that the default option for Failover being the way that it is configured is because normally when it is used, the source site will be down and hence there will be no clash of IP/Name (doh!! I was only doing testing, so had not originally considered this)

Presumably, the way to do a Planned Failover for multiple dependent machines (that would be in a failover plan for normal failover), is to shut them all down at the source site, and then to multi-select them and perform a Planned Failover for the machines in the order that they need to come up at the remote site (as Planned Failover only works sequentially) - is this correct?

[Mike Resseler]

Actually, in a planned failover, VBR will first do a final replication and then shut down the VM at the source site. The exact procedure is explained here: https://helpcenter.veeam.com/docs/backu ... tml?ver=95

If I understand correctly what you want to do, then in the wizard of running a planned failover you can select all the VMs you need from the infrastructure of from the different replication jobs you have: https://helpcenter.veeam.com/docs/backu ... tml?ver=95

PS: Please consider surereplica (if you are using VMware) for replica testing. And in the future certainly have a look at VAO (Veeam Availability Orchestrator) which will give you a lot of power and possibilities for these type of scenarios

[garymansell]

Yes - I have managed to get SureReplica working and can get the machines to come up OK in a Virtual Lab, I guess you are implying that this is the way to test the replication is working without affecting the source machines.

The trouble with Planned Failover of dependent machines is that they are failed over sequentially when you multi-select the VM's in the Wizard and hence this can cause problems with the application (as one will still be at the source site and one at the replica site and it takes time for DNS to propagate the new IP's).

It hence seems to me that to avoid issues with dependent machines when doing a planned failover, it is necessary to shut both the machines down at the source site first (even though they will be shutdown by the planned failover process) and then perform the planned failover in the order that the machines need to boot at the remote site.

[Mike Resseler]

Ah, yes. Understood. And yeah, even if you manually shut it down, you will see that there is still a final replication so that you have the latest information.

Again, in the future VAO will give you the possibilities to really create the workflows the way you want them.

[t481]

@Mike Resseler
Any news on the workflows idea you mentioned?

I need to be able to run a planned failover job which I can hopefully do using a PowerShell script but I need a way to Veeam to update the DNS record of each system as it is failedover. This is possible with a Failover Plan but not with a Planned one.

[veremin]

Failover Plan does not update DNS records of failed over VMs, unless scripts are meant here.

And scripting DNS record update for Failover Plan should not be any different from scripting that for planned failover.

Thanks.