Hyper-v 2012 node - drained - still used as proxy?

[frankive]

While maintaining a 2012 hyper-v failover cluster we drained (paused) a node for maintenance. We rebooted after a few updates during the backup, but to our surprise we saw that one of the job actually had started to use the hyper-v node as a on-host proxy server during this maintenance.
Is this expected behavior?
Is it possible to let veeam detect the maintenacenstate of a hyper-v node and exclude it from jobs during this?

[Vitaliy S.]

Were you trying to backup VMs from another node? I believe you can stop Veeam services and this will prevent jobs from picking up this proxy for data processing.

[frankive]

Yes, it is a 2-node setup. We drained one node to do maintenance. the Vms were running on the other node. Dont know why it picked the drained node for proxy, perhaps because it had a lot of horsepower since it was drained? We are not using direct san mode, just on-host proxy.

Could it perhaps be a ticket in the proxy on the VBR that say "Dont use this proxy while Hyper-v is in drained mode" or something? There must be some kind of event written in the eventless from Microsoft telling that is is in drained mode which the veeam job could check before it picked the right proxy?

[Vitaliy S.]

Hmm...who is the CSV owner in this cluster, the host that you've put into maintenance? Can you please check that?

[frankive]

No, when we drain a hyper-v 2012 nodein a cluster it wil drain all roles, inclusive the csv-ownership.

[frankive]

We have not enabled Cluster Windows Updates which is suppoes to patch the cluster automatically with drain the nodes. I don't feel comfortable with that if Veeam actually will use a drained node as a proxy during the maintenance widows.
The result we experienced was during our maintenance of the node the node rebooted (we did this manually), while Veeam was using it as a proxy (which we see in the logs) and also left a snapshot behind on our Hp LeftHand San.

[Vitaliy S.]

If CSV-ownership was given to another node, then the behavior you see should not be happening. I will ask our QC team to confirm this behavior, meanwhile can you please send your latest job logs to our support team so we could review it as well?

[frankive]

Not sure if I follow this

If CSV-ownership was given to another node, then the behavior you see should be happening

Are you saying that if I drain a node (which is equally that is not part of the cluster for a period) it is expected behavior that Veeam will still use it for proxy? Is that because it has more resources avaiable?

I will open a case and submit the logs for the job. Will the logs actually give you data if a node is drained? Just curious

[Vitaliy S.]

I just missed the word "NOT", so the text should be "not happening". Sorry for that

[frankive]

To be or NOT to be, could sometimes be the question!

[frankive]

caseid #00886139

[Vitaliy S.]

Hi Frank,

I have just received an update from your support engineer, that you've been using these hosts as off-host proxies, even though these hosts belonged to the Failover Cluster. Is there any reason why you've used this configuration?

Thanks!

[frankive]

No, there is no reason for this, it must have been a misconfiguration from the backup operator. We have removed those 2 hyper-v servers from the proxy section so that it will now work in on-host mode which is what we want them to do.
I guess that also explains the behavior we saw while running the upgrades and using the other as a proxy during the upgrade while it was drained.
I will do a new update next month and will certainly let Veeam run during the update to ensure that this is the case.
Thanks!

[Vitaliy S.]

Yes, that's correct, though I'm not sure how you made it work initially In your scenario there should be on-host proxy servers, so that when you put one host into maintenance mode it will not be used as a proxy for the running backup jobs.