Replication recommendations

[glennsantacruz]

I'm evaluating Veeam and ran across this thread regarding replication / WAN / rsync.

@Aaron - if you don't mind, could you please help me to understand what you're doing for replication over your WAN? I think I have a basic idea of your setup, but being new to Veeam, I'm not 100% sure. To help illustrate, let me give you an idea of what I think I understand you're doing:

Site 1 : ESX1 : Linux1 : VM1 -- Veeam is setup with replication job for VM1, to backup target of Linux1
Site 1 connected to Site 2 via 10Mbit Metro Ethernet
Site 2 : Linux2
Linux1 running rsync job to replicate to Linux2

In this situation, you can control bandwidth throttling directly via rsync and also realize other benefits of that tool. What I'm most interested in, however, is what the Linux2 box is doing with the resulting inbound replicant data -- is it writing directly to storage than a set of ESX servers can "see" at Site2, or are you setup such that the Linux2 server itself is your failover machine (using VMware Server instead of ESX)?

Basically we're interested in achieving the same thing, but the Veeam replication piece looks "tricky" for the WAN, with respect to having a "hot" site ready-to-go ( ESX-wise ).

[Gostev]

Glenn, actually this thread talks about syncing backup files to DR site, not about replication actual VMs. For ready-to-go hot site, you need to be using Veeam replication jobs.

I will move this discussion to a separate topic, not to hijack this discussion.

[glennsantacruz]

But the Veeam replication jobs don't allow for WAN throttling... I was hoping to achieve it with another method (instead of going into the routers and trying to traffic-shape between the Veeam server and the "across-the-WAN-target"). I think I would probably benefit from a replication concepts guide (specific to Veeam). Having come from a different model ( esxpress ) to Veeam, I think I'm having some difficulty in reconciling the conceptual differences between the two.

Do you have a sticky thread / FAQ on WAN-based replication? In particular, I've noticed many suggestions related to placement of the Veeam server in relation to pull/push, but I'm still unsure which would make sense for our environment.

Site 1 ( ESX farm ) ---> WAN ---> Site 2 ( ESX farm )

Where Site1 is our production datacenter, and Site 2 is a hot standby ; however, Site 2 also runs a few "production" VMs, so we effectively have bidirectional replication requirements (and want to preserve bandwidth on the WAN). Veeam replication jobs appear to be the winner in terms of keeping the amount of replication data "small"; our original approach involved using hardware-based replication to keep the backups in sync between the two sites ( data domain ). If the Veeam replication can handle this more efficiently, we're definitely interested in that approach...

[Gostev]

Do you have ESX or ESXi, and what version (3.x or 4.0)?

[glennsantacruz]

ESX 4.0u1 (both datacenters) ; both ESX clusters are attached to FC SAN ( HP EVA 4400 ) ; WAN is MPLS 10Mbit

[Gostev]

Glenn, this is basically best case scenario. With ESX4, Veeam Backup can be installed in either Site1 or Site2, there will be very little difference in WAN traffic consumption.

[glennsantacruz]

Best case likely would involve a much larger WAN pipe, but I'm glad to see that this is a good setup for Veeam.

To extend this question a bit further ( specifically for DR purposes ), how would replication work for a site failure? For instance:

1. Site1 (Veeam) --> replicate --> Site2
2. Site1 complete failure
3. Site2 must become "live" ; presumably we can simply:
a) Modify the replica VM properties to match Site2 settings ( network labels, add swap disks, etc. )
b) Start replica VM's manually ( without using Veeam to failover, since there is no Veeam )
c) ensure all works properly

Is this accurate? Or am I missing a vital step? Like maybe install Veeam in Site2, restore Veeam settings (how?), then use Veeam to failover the VMs.

What about failback, after Site1 is recovered from disaster? I would assume we'd set things up in reverse ( Veeam replicate from Site2 to Site1, then failover gracefully off-hours ).

I do realize that each environment will differ with respect to requirements, RPO, RTO, etc. but I would expect the above is fairly "generic" and you may have a best practice guide or document demonstrating the scenario.

Thanks

[Gostev]

Glenn, yes both scenarios you have described will work fine.

This discussion covers a lot of questions about replication/failover:
Veeam Server DR