Replicating from a replica

[marty9876]

Hello All,

I'm trying to replicate from onsite warm replicas to a offsite DR environment however I'm not sure how to set things up. Please review and let me know your thoughts on the below.

Production environment- 2 ESXi 4.1 servers on the same LAN, Veeam Server (#1) on same network (push setup)
ESXi1- hosts production servers
ESXi2- replication target, replicate from ESXi1 to ESXi2 via Veeam with CBT enabled on the Veeam replication jobs.

DR environment- hosted VMware cloud provider with a Veeam Server (#2) on their end (pull setup)

I'd like to have the DR environment with their own Veeam server pull from ESXi2 (replica's). However when CBT was enabled the DR replicas were never updated, assuming this was based on the replica's never being powered on so VMware never saw any changes, and I'm not sure how long the jobs would take with CBT disabled across the WAN.

Any thoughts?

Thanks,
Marty

[foggy]

Hello, Marty!

In case when CBT is disabled, it will take a long time to scan the replica for the changed blocks. Still only changed blocks will be transferred over a WAN link but the whole job will take approximately the same time as a full run.

Thanks!

[Gostev]

Given that your job speed is likely WAN link speed bound, disabling CBT should not have much effect on overall job performance, and will mostly be determined by amount of changed data that needs to be replicated over. Thanks.

[marty9876]

Does Veeam need to pull all the blocks back to the offsite DR Veeam server to tell if they changed? Trying to determine if the duration of the job would be equal to the original replication time.

Will disabling CBT cause the replicas on the DR site to be updated?

Side note: for some reason any Veeam job from my location to DR site is limited to only ~ 3 meg (10 meg pipe). We run 3x jobs at a time to take advantage of the available bandwidth, not sure why

[Gostev]

Offsite Veeam server will determine the changed blocks locally in the DR site (by scanning the whole image), and will then only send changed blocks over WAN.

[marty9876]

Set the offsite Veeam server job up with CBT disabled?

[Gostev]

Oh, I did not read your scenario carefully, sorry. Since this is pull replication over WAN, disabling CBT is bad idea of course.
The only option you have today is pull replicating VMs from ESXi1 with CBT enabled.

v6 should make it possible exactly what you are trying to do, with CBT enabled.

[marty9876]

If CBT is enabled then the replica's at the DR site never get updated. VMware never reports any blocks are changed due tot he VM's never being powered on (they are onsite replica's) is what my guess is.

Not sure what all options this leaves me.

Can you let me know any options you feel there are at this point or speak to how this will be addresses in v6?

[Gostev]

I think you missed the fact that I am talking about using ESXi1 as source, this way CBT will work fine.
That would be the only option available to you today.

Unfortunately, I cannot provide any details on v6 functionality just yet - until the official announcement.

[marty9876]

If ESXi1 is used as a source I'm concerned about the duration of the Veeam jobs. For the initial replication it would take ~ 15days top complete. I'm concerned the snapshot would have issues being committed back against a domain controller and a Exchange server.

[Gostev]

You are 100% correct that this may cause issues, I did not realize it takes so long for you.
Afraid you have to wait for v6 then, in order to do what you want to do.

[marty9876]

Running v5 with CBT disabled would yield the correct results, right? Albeit with likely a similar time frame for the job to complete from the original job to subsequent job.

[Gostev]

Running v5 with CBT disabled would yield the correct results, right?

Not when replicating a replica. CBT does not track external modifications to virtual disks.

[marty9876]

How would you replicate a replica with v5 then? Assuming the replica was never powered on.

[foggy]

Running v5 with CBT disabled would yield the correct results, right? Albeit with likely a similar time frame for the job to complete from the original job to subsequent job.

Yes, that's correct. I think Anton somehow misread your statement.

[Gostev]

That's right, sorry - I misread your post. CBT must be disabled when replicating a replica with v5.

[marty9876]

v6 should make it possible exactly what you are trying to do, with CBT enabled.

Please confirm, v6 should allow pulling from a replica with CBT enabled. Basically, most optimized method for replicating a replica over the WAN once all this are considered. Push versus Pull, ESX versus ESXi etc.

[Gostev]

Actually, v6 replication architecture will no longer have these notions of push/pull or ESX/ESXi. Veeam server location will not matter for efficient replication over WAN. Host type will not matter either. Everything is much nicer there in v6