Dedicated Replication network between Hardware Proxies

[MattM NZ]

Hi

We've two data centres separated by two WAN links - one is layer 3 (Generic traffic) and Layer 2 (Replication). The layer 2 link is a dedicated fibre for site to site replication, SQL log shipping etc.

Our VEEAM Hardware servers/Proxies sit on two each of those two IP networks - one dedicated NIC on a routable layer 3 network, the other NIC on a dedicated layer 2 network with no routes onto or off that connection. Just a flat spanned single IP range between two datacentres. I want to ensure all VEEAM replication network traffic is copied between datacentres on this L2 link only.

HOw is this achieved? Not alot on this in the veeam docs.

Backups of VM's occurs via Direct SAN / Fibre channel straight off storage to the backup proxy.

Cheers

[foggy]

Matt, have you tried specifying preferred networks in Veeam B&R?

[MattM NZ]

Foggy - Great. Never even saw that in the documentation.

Will test.

Despite VM Backups being pulled via Fibre channel, this setting im assuming will only_ apply to Replication traffic ?

[Vitaliy S.]

This setting will apply to all components involved in the data transfer. In your case since source and target IP addresses will be for proxy servers, then these settings should apply to replication traffic only.

[MattM NZ]

Ok this seeeeeeeeeeems to work, However, the regular LAN - non replication - interface traffic soar's during a replication test. The replication Ethernet interface also see's heavy use.

To break it down, the Sending Proxy Replication interface traffic shows significant bandwidth use highlighting bits sent - which is to be expected, However the normal mgmt LAN interface we want to keep all replication traffic away from shows significant bandwidth use as well but specifically receive only traffic.

Are we able to determine what exactly occurs here? I want next to zero VEEAM traffic on my production WAN. Replication is only_ to occur on the replication specific interfaces.

[Vitaliy S.]

The management interface of ESXi host is still going to be used for communication traffic, however all data transfer will go through the connection you have specified to the proxy server. Can you please tell us if you see this load during entire job run or just at the beginning and at the end?

[MattM NZ]

During the entire job. The Repl interface traffic approaches the maximum available on the WAN as to be expected almost for the duration. Its bursty.

The management network experiences high bursts during the entire job as well often as high as the replication network but not as consistent.

[Vitaliy S.]

Can you check with the replication session log when do you see these burst of traffic flowing between proxy servers over the "not preferred" route? The traffic should still be happening, but data will be transmitted between ip addresses you have defined.

[MattM NZ]

Vitaliy, We're running several replication jobs and all of them hammer our layer 3 network, not our layer 2 replication network. I have configured the preferred network setting as our Layer 2 intra-datacentre replication link subnet.

If i observe the NIC's on the destination backup proxy - the Layer 3 generic 20Mbit WAN Interface maxes out. Our 40 Mbit replication network - for dedicated inter-data centre traffic (backup and replication only) is barely utilised.

We have two physical backup proxies, one in each Data centre, both with two NIC's - they can see each other on each NIC. The layer 2 network is a spanned VLAN between two geographic locations ( ie no gateway ). The servers think they are on the same peice of wire. Yet its never utilised. Thoughts ?

[Vitaliy S.]

Hi Matt and welcome back

I would recommend to open a ticket and allow our engineer take a look at the configuration of preferred networks as well as the proxy setup. Without this data, it's hard to say why wrong network is utilized.

Thanks!