Making failover permanent or running on snapshots?

[unsichtbarre]

Hi all,

I was just wondering about opinions/experiences/guidelines regarding when it is time to commit a failover (Make Failover Permanent)? In other words, how much do the snapshots affect the performance of the replica when it goes into full production?

In the past, running on spinning disk, we've made failover permanent after about 24 hours or before the VM went into production. We've since upgraded to all-SSD or auto-tireing SAN and I am wondering about the advisability of leaving several VMs in Failover, managed by Veeam, for a period of several days in production.

Just looking for opinion/conjecture.

THX

[csydas]

In general even for replicas it's not a good idea to leave a snapshot live for more than 72 hours (I think that's the number tossed out by VMware), especially if we're talking about highly transactional VMs. It's not hard to imagine something just eating up the datastore fast and furious and tanking other VMs with it.

I'd personally be wary for anything that's handling a lot of transactions, as I've seen SQL VMs balloon in a matter of 10 some hours during an extended failover test. (Granted, it was 10 SQL servers all at once...but still) Both the Veeam User Guide and VMware are pretty clear about not running on snapshots long term, and I'm inclined to believe them.

So if you're able to, I'd say pick one server that's big enough, give it a failover for a few days and watch it grow, see how it performs, then multiply the resource usage by however many you'd ultimately want to be failed over, and be as grim as you can be in your forecast. DR is probably better to be over prepared for.

[adapterer]

vSphere performance with an open snapshot generally sucks as disk queue depth is limited to 1 by design.

See here - vmware-vsphere-f24/slow-replica-guest-p ... 47176.html - we have not been able to resolve this issue unless we delete open snaps

[unsichtbarre]

Thanks for the input.
The particular scenario is failover to DR during a weekend maintenance window for the prod site. We are leaning towards running in Veeam failover and using Veeam failback as long as the maintenance window goes as planned and we are back at production by Sunday night. Should we go into Monday at DR, we will make failover permanent and initiate reverse replication.