Backup remotes and resume
-
Hi all,
I have a fleet of xcp-ng hosts running xoce. I have one central XOCE instance, with a huge disk array. The intent was to have all of the remote hosts backing up their delta backups to the central location for offsite backing up.
The problem is that if the initial backup is interrupted, then it has to start over because there is no resume. The "official" solution is to bring the initial seed over with a physical disk.
Given that there are multiple hosts that are geographically separated, and that a full backup is done once every week or so, physical transportation is not feasible.
What, if any, are the other solutions here?
On a side note, to be honest, I'm shocked that resume of large transfers is not considered a critical feature, especially given this specific use case.
-
Hi,
It's in our roadmap. It's not the case yet since the VHD export in XCP/XAPI doesn't support resume, since you cannot do anything else than consuming a stream (from the client perspective, eg XO in that case). So with integrated VHD export, there's no way to resume a backup. It's just impossible to "seek" inside the VHD.
However, since the last 6 months, we are progressively switching all backups with NBD protocol, which is allowing to be faster than VHD export because we can download multiple parts at the same time, and obviously seek into the device. That's the foundation of being able to resume backups. We already explained that in multiple release blog post by the way, you should read them if you want to stay tuned about what's going on
In the end, this was not possible technically before switching to NBD, and yes, we always wanted to have this feature and now it's planned
edit: if you use XO from the sources for a business case which is multi-sites, I have to admit I'm a bit shocked on my side that you never considered to support the project
anyway, your feedback is still welcome. -
@olivierlambert
I know I'm speaking out of ignorance here, and I'm sure there's some good technical explanation, but I just can't wrap my head around "can copy data to portable drive and carry it physically, but cannot tar it up and send it in a resumable way". -
@olivierlambert said in Backup remotes and resume:
edit: if you use XO from the sources for a business case which is multi-sites, I have to admit I'm a bit shocked on my side that you never considered to support the project anyway, your feedback is still welcome
We're a small MSP, just starting out, and as soon as we can convince a client of ours to pay for commercial support, you'll get a cheque from us. Which we will physically mail because sending it electronically is not supported
-
@MrNaz Because the way the VHD disk is exported: it's only sequential, and the HTTP connection doesn't support "pause"/"resume" or "seek" into something. The protocol itself doesn't support those "commands" if you like. That's exactly why NBD was created in the first place.
