Force NBD mode to use non-mgmt vmk nic

[Vmagic]

Hi,

Our esx-hosts have 10GbE nics connected to a distributed swich. However, we chose to put the management interface(vmkernel port) on 1GbE nics and standard vswitch, just for the sake of safety/separation. I would like to use NBD mode since the hot add operations can be slow, but then Im hitting the 1GbE limit. Is there a way(register-hack etc) to force Veeam to use a different nic?

[PTide]

Hi Vmagic,

Please tell me - what kind of repo do you use and how distant from the host it is?
Also what about proxies, do you use any?


Thank you.

[Vmagic]

Hi,

2 source proxy which are VMs running on the esx-cluster that I am backuping up VMs from. Running virtual appliance mode at the moment, but as I mentioned, I want to skip the hot-add process since it takes time and we have 10GbE nics in each host that have bandwidth available for backups etc.
Repo is a Hp DL380 G7 connected to two DataON Storage JBODs running storage spaces with tiering and a 100TB volume. Repo is in DR site which is connected to main site via 2 dark fibers (20Gbit bandwidth).

[PTide]

In your configuration you have VBR --> ProxyVM --> ESXi

All backup/replica activity is performed against .vmdks and involves OS freezes, snapshot operations etc. All this magic is managed by ESXi host, which means that NBD commands issued by backup software have to be handled by VMkernel TCP/IP stack, which is strictly tied to your 1Gb NIC...

Considering all of the above, you have only three options:

1) upgrade your management NIC
2) use hot-add mode
3) get another management network on your vDS

Personally I'd stick to hot-add, because if you use NBD you will get an extra network overhead right between your proxy and ESXi.

I agree that hot-adding may take some time, but I'm curious how big is the difference between these modes in your case? Have you already compared the overall backup time for both?

Thank you.

[Vmagic]

Forced a proxy to run NBD mode and looked at a job containing 2 sql-servers running once every hour. Hot-add took 40-50 min, NBD taking approximately 25 minutes.

[PTide]

Ok, I see.

Btw, what is the reason behind your solution to assign management role to a dedicated vSwitch? You've mentioned safety/separation, could you elaborate on that, please?

Thank you.

[Vmagic]

We basically just had a bunch of 1Gbit ports available after we moved from 1Gbit to 10Gbit and thought I might be good to have a bit of separation. If the 10Gbit nic dies, then well still have management. I did not know at the time that Veeam pulled backup-data through this interface. Easiest way is probably to move management to the distributed swtich.

[PTide]

Can you add your 1Gb NICs to distributed switch so they work in team with 10Gb so in case of failure management migrates to 1Gb?

[tsightler]

With NBD mode the proxy connects to the host using the IP or DNS name has provided to it by vCenter. If vCenter knows the host by the management interface then that is the interface the proxy will attempt to connect to. If you want the Veeam traffic to use a specific network then you'll need to get the proxy to use that specific NIC using one of the following two options:

1. Host tables - If your ESXi host are registered with vCenter using DNS names then you can use host table entries on the proxy so that the proxy resolves to the IP address of the secondary NIC.

2. Routing - With this option you place the proxy on the same network with the secondary NIC and then add static routes to the primary IP address of the ESXi host via the secondary IP. You'd need a static route on the proxy for every individual ESXi host. I like this method less than the first, however, it has the advantage that it can work even if the ESXi host are registered with vCenter by IP address.

[PTide]

Btw, there was a similar topic back in 2013, please review.

Thank you.

[Gostev]

Tom is correct. That said, the secondary NIC still must have management interface enabled on it, otherwise NBD will not be possible.

[jdsalasra]

With NBD mode the proxy connects to the host using the IP or DNS name has provided to it by vCenter. If vCenter knows the host by the management interface then that is the interface the proxy will attempt to connect to. If you want the Veeam traffic to use a specific network then you'll need to get the proxy to use that specific NIC using one of the following two options:

1. Host tables - If your ESXi host are registered with vCenter using DNS names then you can use host table entries on the proxy so that the proxy resolves to the IP address of the secondary NIC.

2. Routing - With this option you place the proxy on the same network with the secondary NIC and then add static routes to the primary IP address of the ESXi host via the secondary IP. You'd need a static route on the proxy for every individual ESXi host. I like this method less than the first, however, it has the advantage that it can work even if the ESXi host are registered with vCenter by IP address.

Sorry to bring up a 5 years old post, but im facing the exact same situation here.

Proxy only has 1 NIC under network 192.168.10.x.
My ESXi hosts has 2 vmkernels, one on network 192.128.5.x (vmk0) which is a private lan and cannot be accessed from any other vlan, and it also has a vmkernel under the same vlan as proxy 192.168.10.x (vmk1). I can ping the vmk1 from proxy OK.But when i try to backup vmdks adding hosts from vCenter it fails because it tries to connect to network 192.128.5.x instead of 192.168.10.x. because the esxi are registered by IP on vCenter.

So in that scenario, i would have to go with option 2, but i'm a bit lost as how to create a static route from the proxy to tell the backup to use the vmkernel 1 instead of vmkernel 0 since vCenter automatically points it to the vmk0 as it is the ip which it was registered the host on the vCenter. Any idea? I'm not an expert on networking.