Using Endpoint Backup for Windows Failover Cluster?

[wasc]

Hi,

We have been using Veeam Backup & Replication for a long time - it's an awesome product. However it's biggest weakness in my opinion is backing up a Windows Cluster (for example a Windows File server cluster). This is because the Windows cluster needs shared storage (for example in-guest iscsi initiator). I got excited when I saw that Esxi 5.5 added support for virtual RDM's shared across two Esxi hosts, but it won't allow snapshots in this configuration so Veeam still won't work.

However, I've been wondering whether Veeam Endpoint Backup could assist in this configuration. I installed it as a test on a Windows 2012 R2 File server Failover cluster we have - interestingly Veeam detected the data disk on the 'active' Windows node, but not on the standby cluster node. The potential issue here is that if I configure Veeam Endpoint to backup the active node, and the node subsequently becomes the standby node, then Veeam Endpoint will be unable to backup the disk any longer.

On the standby cluster node, the shared disk is visible in diskmgmt.msc but marked as 'reserved'. However a query using diskpart seems to suggest the disk is available as read only. So, I have two questions.

a) Could the logic of Veeam Endpoint be modified to support backing up the 'reserved' cluster disk on the passive node? The fact DiskPart seems to suggest the disk is read only is promising - it sounds very similar to the Direct San backup method where the disk is attached to the backup node but 'offline' so the backup node doesn't write to it.
or
b) is there some sort of ability to control Veeam Endpoint through the command line? Or better still, have a pre-job script option? I could potentially have a pre-job script which forces the cluster to switch over to the node Veeam Endpoint is running on.


Apologies for the long post, but the way I see it, this is the closest we've been able to get with backing up a Windows Failover Cluster using Veeam

[wasc]

Sorry for bumping my own post, but I've done further testing with this. The good news is that this does work at backing up a Windows File Server cluster.

My Environment
Esxi 5.5 (latest build)
Two Virtual Machines - each Windows 2012 R2 Windows Failover Cluster servers with the file server cluster role configured.
The cluster disk is presented to both virtual machines via in-guest ISCSI initiator.

My Configuration
Veeam Backup and Replication backs up the Virtual Machines "Node1" and "Node2". However it obviously doesn't backup the important file data as that's on the in-guest iscsi so to solve this;

Veeam Endpoint is installed on Node1 and configured to backup at 9pm of just the cluster shared disk (not the OS disk)
Scheduled task runs at 8:58pm and runs a single one-line powershell "Move-ClusterGroup MyFileCluster -Node node1"

This makes sure that when veeam comes to do it's backup, Node1 is the active node, and Node2 is standby. This is because Veeam endpoint doesn't see the disk on the standby node (even though it's visible in disk manager), but will see it on the active node.

This seems to be working which is awesome. I really think Veeam should expand the use case for Endpoint Backup to backup clusters, or release a different version of Endpoint targeted for these servers?

I still do have some problems with this though. The biggest ones are;

- Cannot schedule more than once a day (we'd like hourly)
- Cannot do pre/post scripts. A pre script could run the move-clustergroup powershell command.
- Would be good if Endpoint could backup regardless of whether node is active or passive.

[Mike Resseler]

Hi Alex,

Nice thinking here . I do see some potential issues. What if the cluster fails over (whatever reason, maintenance or server down or ...). In that case you don't have any backups anymore during the time that node 1 is down and you have a potential risk.

Second, why not backup the OS drive? Let's assume it is indeed a serious disaster, you can do a BMR of only the system volume (the data volume is connected to the other node anyway) to get the failed node up and running asap.

And on your other questions. This is v1 and we are reading very carefully what you guys want. I can't promise everything, but we are certainly looking at a lot of things here at the forums and see how far we can go with a stand-alone product

Thanks

Mike

[wasc]

Hi Mike

Good questions. If the cluster fails over, then the backup will also fail as you point out. This is why I have the powershell script which runs to ensure node1 is active before the backup window begins, but if the cluster fails over during the backup window, then the backup will fail. However it would be easy for an engineer to see this is the case and re-run the backup as necessary. If veeam could backup the disk regardless of whether it's the active node or not, that would potentially solve this issue.

Reason for not backing up the OS drive was because we were backing up that separately using Veeam Backup & Replication. This way, in a disaster we could do an instant restore to get the OS back online. However we'd then need to do a traditional Veeam restore to get the file data back for the cluster data disk.

Hope this helps. I really think you can push this as your solution for backing up a cluster in a windows environment as there must be people out there using this sort of setup, and this is really one of the very few features Veeam B & R is lacking.

[Mike Resseler]

Thanks Alex,

Good, useful additional information.

Cheers

Mike

[jcmachadouga]

I have been testing this but I get to a point where Veeam is not able to decide what disk is what and you have to do a DISK MAPPING. Are you having to do a disk mapping on a bare metal restore?

Thanks.

[Dima P.]

Are you having to do a disk mapping on a bare metal restore?

It happens when you are restoring to dissimilar hardware or disk layout was changed prior Bare Metal Recovery. Were you able to succeed in restore?

[jcmachadouga]

I figured it out.

I was loading the drivers for the NICs (to be able to connect to the Veeam repository over the network) but I wasn't loading the other drivers, including the one for Paravirtual SCSI so it wasn't seeing all the RDMs.

It is all working now.

Thanks