MSSQL restores and network bandwidth.

[staun]

We have a problem when we restore a MS SQL database directly to one of our database servers. It consumes all bandwidth on the network, so all virtual servers on that same host become non-responsive.

When we do a normal VM restore to a host, we dont have any problems.

We suspect that there might be a network configution error somewhere, but I dont understand the difference, except that a VM restore is to our HyperV host and a SQL restore is directly to the VM.

right now we are not able to restore if needed unless we got serious network issues.

[staun]

ok, we have narrowed it all down to our firewall and can exclude all other parts in the setup.
(we just tried to put all servers on same network and restore was very fast and no problems at all)

When a SQL restore starts in consumes all possible bandwidth and the CPU on the firewall jumps to 70% and increases until it becomes unresponsive. The same second I stop the restore everything returns to normal again..

Does anyone have their backup-setup on a separate network/ vLan, that have to go through different firewall interfaces and experience similar symptoms.?

[soncscy]

Maybe it's something to do with iscsi traffic? SQL direct restores in some cases use iscsi traffic to mount the disks if I remember right, so maybe your firewall doesn't like the surprise iscsi traffic appearing?

[staun]

@soncscy
Could be. We just tried to move our backup-setup to another firewall with 10Gb interfaces and now we dont see any problems at all. restores are also A LOT faster.
We are planning on moving to this new firewall in a 3-4 weeks, so conclusion right now is to move that migration up a little. So we are able to do restores.

But will absolutely follow up on that iscsi traffic as it might increase or reappear when we have migrated everything.
Thanks