Which first, Replication or Backup?

[chjones]

Hi all,

We have only recently purchased Veeam and I am still in the process or ironing out any kinks. The first immediate goal I have to deliver is the replication of a handful (no more than 10) VMs between our two main datacentres. We currently only have a 16Mb/sec link between the two sites but by seeding the VMs I have been able to get replication working and seems to be going ok.

However, I have a query about also backing up those same VMs. The thinking in my head was to backup those same VMs in their own Backup Job locally at the source site using a physical proxy with a LUN from our HP EVA8100 presented over the 4Gb fibre, and then run (now that it's so simple in 6.5) the Replication Job for those same VMs as a chained task after the backup finishes. I figured I would leave other VMs that arent replicated in their own jobs so I don't have to wait for them to complete to fire off the replication.

My logic was that the backup is local and over the SAN so it will complete a great deal quicker than the replication, so I'd get the backup done and out of the way immediately and then the rest of the night can be dedicated to the replication when there is little use on our WAN.

What I have noticed, however, is that the backup runs perfectly normally using CBT but the replication then runs and obviously can't use CBT as it's already been used for the backup and that history of changed blocks is now cleared from the VM (it is cleared after a successful backup, right?). The replication falls back to doing a block-level compare of the source and destination VMs which consumes many hours. Transferring the compressed data usually isn't an issue even though our WAN isnt the fastest, it's the block-level compare and how long it takes that is killing me.

I know the logical solution is to run the replication job first but I am concerned if there is problem or bottleneck with the WAN that the replication job will just keep running and never get that backup done. At least if I have a backup first then I have some kind of safety net.

Does anyone have any experience with replicating and backing up the same VM or some advice/recommendations they can share?

Thanks,

Chris

[chjones]

I may have just solved my own issue but I'm hoping someone can validate it for me before I implement something stupid

What if I was to disable CBT processing on the backup job but leave it enabled on the replication job? Would this solve my issue? Since the backups are local and all directly over the fibre would there be any downside other than the backup taking a little longer to complete?

This way the backup could run and would compare each block, but since its all SAN attached it shouldn't be too bad. Then the replication can use CBT to quickly process the VM and only copy those blocks.

Does this sound feasible?

Chris

[Gostev]

history of changed blocks is now cleared from the VM (it is cleared after a successful backup, right?)

Wrong, this is not how CBT works. Please read up the sticky FAQ topic on CBT.

[dellock6]

I wrote it in another thread but I think is worth copying it here too:

Think about CBT as a log of changed block, each with its position and timestamp. When a software reads CBT informations, it searches for a desired timestamp (in Veeam is the time of its previous execution), and read which blocks have changed since then.
Another software (could be two different backup jobs) can read the same log and search for another timestamp. That's why it can support several backups without any interference.

Luca.

[chjones]

Thanks Luca, that clears it up for me. Apologies for the confusion. So I guess in that case it makes no matter if both have CBT enabled, that's great to know, thank you.