Veeam using NBD for certain datastores

[maverick964_uk]

We have veeam 6.5 update3.

The environment was built around a "management" set of datastores.
Veeam lives in this region. Backups and restores use HOTADD no problem.

New nonprod and prod datastores have now been created however backups and restores are using NBD and not HOTADD.

The veeam backup files show this:

Detecting hotadd access level....
Datastore [xxxxx130] does not belong to proxy.

I have updated the backup proxy config in veeam and set to manual and added all datastores. Still get same problem.

Any ideas???

thanks!

[luckyinfil]

If your veeam proxy server is a VM, the esxi host that it resides on does not have access to the production datastores, which is why it cannot hotadd.

[veremin]

Hi, Mike.

Yep, in order to use Hot Add method, the host that runs proxy server, should have an access to the datastore where VMs to backup/replicate are located. Otherwise, the Hot-Add won’t be possible.

Can you confirm that this isn’t the issue?

Thanks.

[maverick964_uk]

Hi,

thanks for the reply. This was my thinking however as its all under control of one vcenter therefore I thought it would be able to do it via ESXi...

we have one vcenter host but we have a set of HP blades for prod, set for non-prod and set for management. Veeam is in management.
Same storage array....

They will be on seperate networks too.

So, the question is, what connectivity is needed? IP connectivity between the blades or do we just present the non-prod and prod storage to the management blades?

Over time we will probably adopt a physical proxy server but we need to move to hotadd for now.

We have found that the backups doing work over nbd however restores are very slow then fail with "lost connection"

[luckyinfil]

If you have separate clusters, an easy option without tinkering with the LUN masking is to just create new VM proxy servers in each of the different clusters and in the backup job, tell it to only use these proxies.

[chrisdearden]

just present the non-prod and prod storage to the management blades

this should do the trick. The other option is to deploy proxy VM's on your compute cluster , but you would need to establish a network route back to the repository / backup sever.

plan B is a bit more scale out. but requires a few more change requests.

[maverick964_uk]

Thanks for the replies.

Yes i expected we would possibly need additional proxies for the prod and non prod clusters.
For the moment I would prefer the storage presentations to the management cluster.

Longer term a physical proxy will handle it all so we therefore just present all datastore luns to it.

[maverick964_uk]

until we are able to sort out the LUN presentations, what we find is that the backups work (bit slow..) but all restores fail.
It would seem that the restores are really slow....5mb/s...

after about 30-60mins the restore fails with this.

any thoughts on this?
My initial presumption is that its timing out somewhere as the restore gets quite slow....

Restore only works if we recover the VM to the management cluster (same cluster as veeam server) the vmotion it to the prod or non-prod cluster...

[Vitaliy S.]

Hi Mike,

Kindly please avoid posting log snippets as it is requested by our forum rules. With all technical issues it's better to contact our support team, as people behind these forums are not able to parse/read debug logs.

Thanks!

[maverick964_uk]

yes I understand that but its a small bit of information to try and give someone an idea of what we are seeing......hopefully from there some useful suggestions would come out...

[maverick964_uk]

we get a connection error in the backup logs. Is veeam expecting regular contact from the restore process?
I'm not sure why it runs for 20-60 mins then we get a connection error and the restore fails.

Are there timeouts we can increase?
thoughts?

[Vitaliy S.]

I'm not aware of any reg keys or timeouts for restore operation, let's see what our support team says after reviewing your log files.