WAN replication over a dedicated link

[mattmnz]

We've this ongoing issue....

Two Datacentres with two ESX Clusters separated by layer 3 routed WAN link (20 Mbps) & a layer 2 spanned VLAN (100 Mbps) for replication purposes. Each cluster has a single physical VEEAM Backup Proxy/WAN Accelerator host. These hosts have two NIC's. One physically connected to the layer 2 link, the other to the layer 3 link ( Gateway/routed ).

We have VEEAM configured for Replication link subnet preference for replication purposes & have defined the layer 3 IP net properties with throttling.

The issues are

1. Throttling occurs on the layer 2 link as well as the layer 3 link. When we remove throttling...
2. Layer 3 pipe gets absolutely HAMMERED with VEEAM traffic to the point of total congestion. Layer 2 link is utilized in a mirror capacity to that of the Layer 3 link. If it can hit 10 Mbps on L3, it does so on L2.

Essentially, Replication traffic does not stick exclusively to the layer 2 link as required. I've been at this problem for 6 months and have tried every trick in the book. The host file on each VEEAM Proxy has its partner configured via its layer 2 IP.

Every replication job when it's kicked off totally destroys our layer 3 network and seems to trickle across the layer 2 link.

Thoughts and suggestions welcome before i go postal.

[foggy]

Hi Matt, I'd let our support engineers to review your setup more closely.

[mattmnz]

Solution::

In our case, if you have a dedicated Ethernet interface for Replication/WAN Acceleration - your WAN Accel servers need to be present in the VEEAM configuration with their Replication IP Addresses only.

It would appear that the "preference for replication subnet" feature doesn't really work - at least not in our scenario.

It might pay in this regard to also have the Proxy Hosts present with the replication IP as well, if that is indeed your network topology. Its worth noting we have all roles on physical servers. We have not split out the VEEAM server/Cache/Proxy services into different hosts.

[mattmnz]

Oh, One more thing, The network traffic seen on the "mgmt" L3 interface (that is - NOT the replication interface) appears to be one veeam service talking to the other via loopback. You may still see huge amounts of traffic in/out of this interface but replication traffic should only traverse the replication interface if the accelerators are present with their replication IP's as mentioned above.