Live (hot) system backup

it’s a controversial topic … there are a lot of opinions out there … I would love to get some opinions from manjaro users

the most common opinions say that backing up a live system (using the system itself) is foolhardy because files may be changed while the backup is running

and yet, some developers of backup software (who probably really do know what they are talking about) say that it can be done under the right conditions

fsarchiver (my favourite) fsarchiver.org_live-backup - hot backup can be done if there is no risk of inconsistency (if read/write operations are not going on)

borg (very cool) borgbackup.readthedocs.io_en_stable_faq.html#can-i-backup-my-root-partition-with-borg - backing up your entire root partition works just fine …

so, I’m thinking that if nothing in particular is happening on my ordinary desktop system, there should be no risk in doing a live backup

what do people think about this?

ps modified the links in the post - forum won’t allow them for some reason

Of course you shouldn’t perform IO operations while backing up. Besides that it works fine.

yeah i any case don’t do hot backups while upgrading your system.

I use timeshift in rsync mode. It does snapshots and can be done also while upgrading, but i would not recommend this.

If you want to do a fresh install, i would rather recommend to clone the whole partition as img or as read-only squash file in a live environment.

There is an easy remedy for this problem: create a snapshot and backup from the snapshot. The running system may change files but the snapshot will stay the same and consistent.

Psssh, I’ve done dd backups from a live machine before. I don’t remember if I ended up using it (the backed-up disk image), but I think I did.

There was a lot of seat-of-the-pants admin stuff going on back then. I’m glad I’m out of that job.

Thanks everyone for the answers - glad to hear that a hot backup works ok. A backup plan should be easy, otherwise it trends to get ignored. It gets a bit tedious having to shut down, attach the backup drive, boot into another system, do the backup …

good advice, thanks

I know about timeshift for snapshots - other methods like snapper or squashfs would need btrfs or LVM systems, right?

Now that’s living on the edge!

borg is great. It’s fast, takes little space and images are mountable. Once you get commands on script, it’s easy to manage.

1 Like

squashfs is a file system and can’t use btrfs (another file system).
Yes - all snapshotting is either done via file system (btrfs, zfs) or lvm.
Timeshift/snapper/whatever would only utilize those or cannot provide “real” snapshots.

yes it looks excellent - I’m just working on getting my head around its pattern matching system

OK, I get that now. I think for my situation “not real” snapshots should be OK

There’s a man page just for patterns: man borg-patterns

yes, it’s a bit of a different system but seems quite flexible