Off-site backup - how do you do it?

[grl]

As a bigger change in our infrastructure is right around the corner we're also thinking about new strategies for offsite-backups.
For now we use a dark fiber to another site - so we had no worries with bandwidth and backup times. Just deploy a big enough storage on that remote site and see your data fly there via a 1GBit/s dedicated, rocksolid and stable connection.

Unfortunately we're going to loose that great connection - so we have to deal with a 30MBit/s "normal" internet connection in the future - and have to share that with the "normal" internet traffic we need.
We've a bunch of VMs and a file server that need about 1TB backup space all together right now - that is a bit much for that low bandwidth to backup it every night.

I'd love to know how proxmox ve users around the world handle that.
Cloud-storage? With which protocol? Do you use public services or do you host your own infrastructure somewhere in a datacenter?
Tapes? Where do you store them? Do you have SOPs for the loss of a tape?

regards
Lukas

 

[guletz]

I'd love to know how proxmox ve users around the world handle that.

... me too

Some thoughts from my "world" -
- it is not so important if you have 1 Tb or sevaral Tb, but the new increment data that you have (one thing is to add 100 Mb/day as new data and other thing is to add 200 Gb/day)
- if you use zfs it is very effective to use pve-zsync from main DC to a remote DC (it is incremental), something similar could be in case of ceph (as I read on this forum )
- you can use also a cloud provider if you trust them
- tapes are very good but you need a lot of money for them

Good luck/ Bafta

 

[James Crook]

We have containers so use Borg as a backup method.

 

[LnxBil]

As @guletz already described, ZFS is a bit win. We just snapshot our servers regularly and do incremental send/receive to our off-site-backup server. We also do external backups from the backup site via single hard disk ZFS pools and do incremental send/receive backups as well.

 

[guletz]

Hi,

As a side note, zfs send/recive is the best tool, because zfs knows what are new blocks of data. Any other method like borg are using the file-system information, so it will need to read ALL metdata and also many files. And if you have millions of files, this will take a lot of time(find the files to copy, open file, transsfer file, close file = a lot of files op).

Good luck /Bafta!