CDP replicas and power outages

[mwhite]

Hi All,
I am configuring CDP for a customer that has 2 clusters with 3 hosts each.
I'm trying to understand how we should handle the below scenario.

Veeam CDP sends replicas from Cluster1 to Cluster2.
Cluster 1 suffers a complete outage (eg sudden power failure).
The outage will be extended so a failover plan is triggered to bring up the replicas in Cluster2.
Power is restored to Cluster1, the hosts boot and as part of HA the running VM's will be booted.
Both the original and replica machines are running which will trash everything.

My questions are;
Does Veeam account for this and will it proactively stop/block the boot of the original VM?
Are there any settings we should be configuring on the cluster to mitigate this (only thing I can think of is to disable the host failure action but that will prevent VM restart on a single host failure).

[HannesK]

Hello,

Both the original and replica machines are running which will trash everything.

who says so?

If you fail over, then the replication does not work anymore. I recommend to use a test machine and play around with failovers.

Best regards,
Hannes

[mwhite]

who says so?

My wording is probably poor, but the main issue I'm concerned with is after the power outage HA will power on the running VM's. Which will result in the original and replica running at the same time, while Veeam won't replicate we will still end up with the issue of both machines running, potentially having users connect to both.

How does Veeam recommend users handle this or does it have smarts built into it already?

[HannesK]

Hello,
if the DR machines are powered on because of a (manual) failover triggered in Veeam and at the same time something starts the production machines, then this would cause issues, yes (assuming the network configuration allows to create issues). Veeam cannot prevent humans or VMware HA from powering on production machines.

The main question is, how such a situation can happen. If we talk about power outages, one manually has to decide to declare a failover. If that person is unable to prevent VMware HA to kick-in, then it's a problem. That has to be solved during "boot up" of the datacenter, because one (Veeam) cannot configure a powered off device.

Best regards,
Hannes

[mwhite]

Hi HannesK,
So long as the default action for 'Host Failure' is configured to restart VMs, it's not really feasible for someone to intervene in the boot process. I was hoping that since Veeam talks to Vcenter it would see the hosts come back and then either prevent or power off the VM's as HA tried to restart them since it would be aware that it has a replica running but if that's not the case then we will need to figure out what the process would be.
The only other options I can think of is to physically prevent the cluster from powering on automatically or disable the host failure action but that then means that a single host failure is not automatically remediated.

[HannesK]

Hello,
independent from CDP, the same would happen with restoring a VM or snapshot-based replication.

As far as I remember, vSphere HA works without VCenter. That means the VMs might start even before VCenter is available for Veeam (because VCenter start takes long...). I don't see even a theoretic chance how one could influence something there from our side.

Best regards,
Hannes