Backup Copy: Never transfer a full backup

[Regnor]

Hi @all!
I should probably check that in a lab and look closer at the Backup Copy feature but perhaps the community can help me understand this one

Scenario:
Backup Copy is used to move the daily backup to an offsite repository via a tiny WAN connection. The first full backup will be manually moved with a USB disk to the target repository so that it can be used for seeding/mapping.
The WAN connection will be too slow to ever successfully copy a full backup offsite and the primary backup job will create full backups from time to time, so the backup chain will also be recreated.

So, will it be possible to do a forever incremental Backup Copy job without the need to copy over a full backup ever again?
And if so, where are the limitations or risks which could force the copy of a full backup?

[dellock6]

Hi,
backup copy job is by design a forever incremental activity. It does not move a daily backup offsite as you said, rather it synthetically creates a restore point using a primary job as a source for its data. So it does not matter if the source job has a weekly full.
As long as the backup copy job chain is kept in place, incrementals at the beginning of the chain will be merged with the full, but this operation is done by the repository at the target side, so again no need for huge transfers over the wan.

If the bandwidth is a huge issue, you should be careful to not restart the chain, or also avoid operations like defrag into VMs which would lead to large amounts of changed blocks in the primary job, which will then moved into the backup copy.

[Regnor]

Hi Luca,

thanks for your reply.
Could you please also explain to me what exactly you mean with chain and how I would not restart it?

[foggy]

Backup chain is a set of files containing VM restore points - e.g. first full (VBK) and subsequent increments (VIB). By restarting it Luca means any action that would cause entire VM data to be transferred over a slow link (can be a consequence of changed VM ID, for example, or job re-creation).

[Regnor]

Regarding the VM ID; if I restore a VM, does this mean that I can't map it to an existing Backup Copy?

[foggy]

If you'd add the VM restored to a new location to existing job, it will be copied in its entirety.

[Regnor]

Ok I understand that.
Both of you many thanks for the great explanation!

[ITP-Stan]

Make sure you have a repo proxy server on the target side.

I have tested with a SMB repo available via WAN (simple cloud storage), but the merging is really slow because the server on the source side does the merging over the tiny WAN connection.

[foggy]

Make sure you have a repo proxy server on the target side.

Right, called gateway server in the repository settings.

[Regnor]

Make sure you have a repo proxy server on the target side.

I have tested with a SMB repo available via WAN (simple cloud storage), but the merging is really slow because the server on the source side does the merging over the tiny WAN connection.

Of course I would install a repository server at the central site, thanks for mentioning that.

I'm currently doing some tests in a lab, with a virtual 5Mbit connection
I'll report back if I experience any problems.