safety and speed of concurrent backups to different repos

[madfusker]

Borg 1.2.7
Main Source - on premise QNAP NAS (6-core Intel(R) Core(TM) i5-8400T CPU @ 1.70GHz)
Target1 - on premise QNAP connected by LAN (typically pretty fast backup)
Target2 - offsite pi on a rasberry pi4 (can take a little more time over 20M connection)
Target3 - offsite pi on a rasberry pi4 (can take a little more time over 20M connection)
All 3 repos are their own repo, as it should be.

I've been using Borg for 4-years now and it's been fantastic! Thanks for all the work on this project!

Typically I have been calling backups sequentially in order via script, which takes a pretty good amount of time to run. During a long backup, if cron rolls back around to the next day, it checks if borg is already running (if /opt/bin/pgrep -x "borg" > /dev/null), and if so it stops the second run from kicking off, then tries again the next day. In each script I pass via variable:
BORG_REPO - Obviously different for all 3 repos
BORG_PASSPHRASE - Same for all 3 repos
BORG_RSH - Same for all 3 repos

The reddit post below is a bit old so I am checking-in to see if there are any worries to try and call all 3 backups at once with Borg 1.2.7, and if that is the right way to go when I have 3 backups to do? When backups are running there isn't really too much disk i/o, CPU tax, or external internet i/o on the Source NAS, so it makes me think that I might be able to minimize my total backup time to 3 locations each night if I can run 3 jobs concurrently with no worries of data corruption. To do this I suppose I would have to call 3 scripts at the same time, which are nearly identical, except for repo location.

Thanks for amazing work on Borg!

https://www.reddit.com/r/BorgBackup/comments/txnvif/how_to_run_multiple_backups_at_the_same_time/

[infectormp]

https://borgbackup.readthedocs.io/en/stable/faq.html#can-i-copy-or-synchronize-my-repo-to-another-location

[madfusker]

ya, saw the docs, but it's not clear if "at the same time" is safe. The docs say don't replicate the same repo, instead do what I am doing described above. Nothing about if you can run 3 instances of borg at the same time on the source, with all 3 borgs pushing data to 3 diff repos.

[infectormp]

The docs say "If you want to have redundant backup repositories (preferably at separate locations) ... This will create distinct repositories (separate repo ID, separate keys)". So, you can run borg at same time.

[ThomasWaldmann]

safety: yeah, running concurrent backups to different repos is safe. borg differentiates repos by their repo ID and chooses respective local cache directories to keep them apart.

performance: if your hardware is capable enough, concurrent backups to different repos should be faster than sequential ones. currently, that's the only way to use more than a single core with borg client(s).

you can even try to run concurrent backups to the same repo (e.g. using a long --lock-wait time), but borg with then serialise them via the repo lock. thus: safe, but not faster.