I run Borg nightly, backing up the majority of the data on my boot disk, incl docker volumes and config + a few extra folders.

Each individual archive is around 550gb, but because of the de-duplication and compression it’s only ~800mb of new data each day taking around 3min to complete the backup.

Borgs de-duplication is honestly incredible. I keep 7 daily backups, 3 weekly, 11 monthly, then one for each year beyond that. The 21 historical backups I have right now RAW would be 10.98tb of data. After de-duplication and compression it only takes up 407.98gb on disk.

With that kind of space savings, I see no reason not to keep such frequent backups. Hell, the whole archive takes up less space than one copy of the original data.

Using Kopia, backups are made multiple times per day to Google drive. Only changes are transferred.

Configurations are backed up once per week and manually, stored 4 weeks. Websites and NextCloud data is backed up every hour and stored for a year (although I’m doing this only 7 months now).

Kopia is magic, recommended!

And equally important, how do you do your backups? What system and to where?

Backup all of my proxmox-LXCs/VMs to a proxmox backup server every night + sync these backups to another pbs in another town. A second proxmox backup every noon to my nas. (i know, 3-2-1 rule is not reached…)

I tried Kopia but it was unstable and janky, so now it’s whenever I remember to manually run a bunch of rsync. I backup my desktop to cold storage on the first of the month, so I should get in the habit of backing up my server to the NAS then also.

No backup for my media. Only redundacy.

For my nextcloud data, anytime i made major changes.

Boils down to how much are you willing to lose? Personally I do weekly

Weekly full backup, nightly incremental

rsync from ZFS to an off-site unraid every 24 hours 5 times a week. on the sixth day it does a checksum based rsync which obviously means more stress so only do it once a week. the seventh day is reserved for ZFS scrubbing every two weeks.

Depends on the system but weekly at least

Thanks for reminding me to validate.

Daily here also.

I classify the data according to its importance (gold, silver, bronze, ephemeral). The regularity of the zfs snapshots (15 minutes to several hours) and their retention time (days to years) on the server depends on this. I then send the more important data that I cannot restore or can only restore with great effort (gold and silver) to another server once a day. For bronze, the zfs snapshots and a few days of storage time on the server are enough for me, as it is usually data that I can restore (build artifacts or similar) or is simply not that important. Ephemeral is for unimportant data such as caches or pipelines.

Daily toward all my three locations:

• local on the server
• in-house but on a different device
• offsite

But not all three destinations backup the same amount of data due to storage limitations.

Backups???

I use Duplicati for my backups, and have backup retention set up like this:

Save one backup each day for the past week, then save one each week for the past month, then save one each month for the past year.

That way I have granual backups for anything recent, and the further back in the past you go the less frequent the backups are to save space