Switching from push to pull replication

[romwarrior]

In preparation for taking my replicas offsite I have been using my local Veeam installation for replication to a local ESXi host. Now I have taken that host offsite and have tried a couple push replications, but need to switch to pull replications in an attempt to reduce job times. I am installing Veeam at the target. If I create new replication jobs using the same vCenter and point it to the same host and datastore as the jobs at the source, will it just figure it out and transfer only the changes?

[romwarrior]

I went ahead and tested a job and it failed since the vmx file already existed in the destination. So is there a way to accomplish what I am trying to do?

[Vitaliy S.]

If I create new replication jobs using the same vCenter and point it to the same host and datastore as the jobs at the source, will it just figure it out and transfer only the changes?

No, you need to relocate Veeam SQL database which stores all job configuration to the offsite location to continue transferring changes only. You cannot "map" new replication job to the VMs already replicated by another job, so if you try to create a new job it will try to replicate a VM from scratch.

Here is the procedure that should be followed: How to move a job from one server to another w/o messing

[romwarrior]

OK, thanks. One other quick question. What about seeding replicas when doing it in a pull configuration like this. Any way to do that with removable storage?

[Gostev]

Unfortunately, seeding is not possible with pull replication deployment today.
However, the good new is, v6 will get rid of this push/pull stuff, and seeding will always be possible.

[romwarrior]

I guess I better not create any new VM's until v6 comes out, then!

Forgive me for another question, but I am just trying to get a handle on this. When you do a pull replication, is the only data that is transferred going to be the data that is needed to create the latest VRB file? So if the latest VRB file is 900 MB then there was 900 MB of data transferred over the WAN? I'm trying to figure out a speed discrepancy between regular (non-Veeam) file copy jobs between sites and replication jobs. Right now, I get about 8 Mbps (on a 10 Mbps line) on file copy operations whereas pull replication jobs seem to be at around 1.5 Mbps, if my assumptions above are correct.

[Gostev]

No, VRB has little to deal with what is transferred during pull replication. What is transferred, is changed blocks data. While VRB contains data from block which were replaced in replica VMDK by the incoming blocks. On top of that, the data in VRB file is deduplicated and compressed (unlike incoming data, which is going over WAN uncompressed).

[romwarrior]

Thanks so much for the info. So is the "Data Size" info on the replica properties page (rather than the "Restore Point Size") a better indicator of the amount of data transferred during a replica job?

[Gostev]

No, completely unrelated. There is simply no such indicator in v5. But we have added one to v6.

[romwarrior]

OK. Thanks again for all the info.