Direct Connection With Crossover Cable

[seizurebot]

Hi all,

I am a long time user of B&R, my setup is pretty simple, I only have two physical machines:

I have one HP DL380 Gen8 running vSphere 6.5 using HP D2600 direct attached storage and one HP DL380 Gen7 running Win Server 2019 which has its own HP M60 DAS. The Gen7 has B&R server installed and it is also being used as the backup proxy.

I have been running network NBD backups to backup the entire vmware host on a weekly basis. Veeam has been working great for my needs but my gigabit network equipment is no longer fast enough for the amount of data I am moving. Before investing in a new router and switch I would like to purchase 2x 10GBE network adapters (one for each machine) and connect them directly with a crossover cable.

Will this work? What do I need to configure on each end? Do I need a crossover ethernet cable or will any cable do?

Appreciate any insights and advice,
TJ

[tyler.jurgens]

You can directly connect two 10 GBE SFP+ NICs together without a problem. There is no such thing as a cross-over cable in that situation. You must manually assign IP addresses to the NICs on each server.

Otherwise, you should be fine.

[chris.childerhose]

Also, I have another question: Which server in your configuration is the Repository server? I'm just asking as it could lead to slightly different configuration suggestions, but using the two 10GB NICs is doable as I have done this in the past but need static IPs like Tyler mentioned.

[seizurebot]

Appreciate the replies:)

The Gen7 server is also the repo - B&R server, proxy, repo, SQL and all other roles installed directly in Windows Server, no VM.

I left some details out of my original question:

I have one other machine being backed up, a Win 10 desktop PC, it has the veeam agent for windows installed. I still need that backup to go through the gigabit network. I won't be removing any of the gigabit ethernet setup, I still need to access both machines via normal LAN network and I need the Gen8 and its VMs to continue to function as-is over my gigabit LAN.

It sounds like I just need to pop one 10GB ethernet NIC into each machine, hard code the IP (I'm going to use something outside of my normal range like 10.10.0.2 for one machine and 10.10.0.3 for the other - rest of network is 192.168.xx.xx).

Connect both machines with a straight cat6/cat7 and that's it? Veeam will know which nic to use for which situation?

Another idea I've tried in the past but did not have much luck: do away with the Gen7 entirely, convert the Windows Server on Gen7 to a VM, import that VM into esxi on the Gen8, but then I'm not sure how to connect the MSA60 DAS in a way that gives the new veeam VM exclusive access. I know about device passthrough; I could pass-though a raid card or HBA to the veeam VM, but it starts to get too complicated too quickly.

Is there a smarter/better way to go about this? I have a large amount of data, about 80TB with 1-2TB of changes being copied over each week. The deltas take about 5 hours and a full takes me aprox 2 weeks! I'm hoping to move data 2-3x faster with 10GBE, my slow rusty disks are probably the next bottleneck.

[chris.childerhose]

You may need to define the networks within the Veeam BR console under the Network Traffic Rules on top of everything else and then test to ensure the traffic flows through the 10GB network.

[seizurebot]

My 2x 10GBE nics arrived today, I updated the firmware on both but now I'm stuck as I don't know what/where to configure for the network settings.

Rest of the network is on 192.168.1.xx - should I set static IPs in this same subnet? Or do I use something else like 10.100.100.xx? What about the gateway, leave it blank or use the 192.168.1.1 as gateway and DNS?

In VMWARE do I add the new nic as a new uplink for the existing vswitch? Currently I have only one vswitch for VM Network and Management network.

Sorry for the newbie questions

[chris.childerhose]

If you need just backup traffic to go over those NICs then I would configure a different IP range to be used. I might create a new vSwitch in VMware to add the 10GB NICs to so that it is separate from the other traffic. Unless you want to permanently use 10GB for everything.

[seizurebot]

I would love to but can't use the 10GB for everything - I don't have 10GB switching or routing.

I created a new vswitch in vmware, added both ports as uplink (it's a 2 port NIC), then I created a new port group and a new vmkernel NIC. The new port group was given the 'Management' service in vmware. I hard coded the new port group IP as 10.100.100.5. Over on the other machine I set a static IP of 10.100.100.2 in windows. One port is connected but vmware is saying the link is down for both connections. I don't understand what I am doing wrong here.

I'm using a HP branded direct connect cable with the sfp+ modules buil-in to the cable. The NICS, the cable, are all used cheap parts from ebay. I'm wondering if I have a bad component.

[Outlaw]

1) You must check network connectivity between 2 hosts(for example ping)
If you have difficulty understanding how to do this, hire an system administrator

[seizurebot]

@outlaw - that is super helpful, I am glad you took the time to post this insightful and detailed response.

@everyone else:

I could not for the life of me get the direct connection working, VMWARE specifically was saying 'link down' in all configurations. I ended up buying a cheap Aruba switch that has 4x 10GB SFP+ ports, VMWARE was STILL reporting link down - frustrating. It turns out that a reboot of the vmware host was needed for the connection to come online. Not sure if that was a software or a hardware quirk, either way this is up and running now with a 10GB switch.

My actual backup speeds increased by less than was I was hoping - it seems to max out around 1.7gbs, a little over 2x what I was getting before. Copying a file from A to B I am getting closer to 3gbs sustained, but I know veeam is doing dedupe, encryption and other processing which adds overhead.

I think we can close this one out as resolved, please and thank you.

[chris.childerhose]

Glad you were able to figure it out. Who knew a reboot would be the ending solution.