Backups are skipped when syncing PBS to another PBS

[powersupport]

Hi,

When we try syncing PBS to another PBS some of the backups are skipped. Please refer to the attached warning.

==
2022-08-02T16:47:30+05:30: skipped: 73 snapshot(s) (2021-12-26T17:30:02Z .. 2022-07-31T13:30:02Z) older than the newest local snapshot
==

May I know if there is any way to resolve this issue?

Please advise.

Regards,

 

[fabian]

this is currently not possible, the sync uses the last local timestamp to determine the cut-off for syncing. if you want to sync all snapshots, you can sync into a namespace not containing that group (it will still share the chunk store, so already existing chunks won't need to be re-downloaded).

 

[cheesecake]

@fabian, is there any news when this will be possible? it would be really great with this feature, gladly configurable too.

best cheeseck

 

[fabian]

no ETA at the moment. initial patches were written, the bug to track progress is https://bugzilla.proxmox.com/show_bug.cgi?id=3786 (there are two aspects mentioned there - both syncing "gaps"/missing snapshots, and re-syncing known-corrupt snapshots from a hopefully still good remote)

 

[santacruzskim]

+1 here.
I recently set up an offsite backup server ("pbs2") to sync with my home backup server ("pbs1") and on my first sync with "pbs1" I choose to only sync the most recent snapshots via the "transfer last <n>" option. I did this to only transfer a subset of data first to confirm the connection, monitor transfer speed, make sure nothing horrible was happening, etc. before sending off a full sync of multiple TB of data. The "transfer last 1' job ran successfully, but when I flipped it to "last 3" or even "all" it would just report that all other backups were older than the latest (which had already been transferred earlier) so they were skipped.

I came across a workaround used in similar scenarios of creating a new namespace in the destination (pbs2) datastore and editing the sync job's "Local Namespace" to point to the newly created one and it's running now, backing up the remaining chunks.

This is obviously not ideal. If I say I want "all" backups, it shouldn't override that simply because there are newer backups - in the world of backups, newer isn't always better - and using namepaces in this way is misusing the feature to work around the real problem.

Hopefully this use case will help guide any future development in patching this (tiny) hole in this great product.

 

[rschinkoeth]

i came across this today too and i gotta say: what the actual fuck. i almost deleted old backups that i kept around for good reason. how the **** can the sync skip old backups without any possibility to config this? i had synced from my nvme first thinking "hey then its faster than syncing everything from the slow spinning disks", and now i almost "lost" (ofc i have offsite backup, but holy fuck what a hassle this would have been to have to realize this after deleting the other local storage due to space constraints) tons of old valuable backups!

sorry my wording here, but the way this is implemented is so absurdly counter intuitive! and dont even get me started about protection flags not getting synced, or comments! if i type a comment on my nvme datastore, why the flipping hell would i not want the comment to also be there when i sync the backups to my nas spinning disk datastore? in another thread i read thats considered a feature! ngl, but how one can call this a feature is absolutely beyond me. this should be the DEFAULT, with a toggle to turn it off for the 5 people on the planet that want to remove their comments!

the more i use this, the more i realize: if you don't manage this knowing 100% what you are doing you can lose data so obscenely easy!

i really really love proxmox, but the backup server? tbh i wouldnt even say it needs more "features" (as in being able to sync protected flags, it syncing all old backups by default etc), it needs FIXING! nobody expects this behaviour! to be frank: i may have lost data due to this without even realizing up until lately! and i don't even wanna start to think about it.

and then theres also no proper way to cleanup old backups of vms that arent in use anymore. i am legit writing my own python tool around the api in order to be able to cleanup stuff thats unused, but where i wanna keep around a certain amount of backups just in case. for a tool thats this mature, and given the quality i have come to expect and gotten used to from pve, the way pbs is working is just a sad joke.

you really gotta step up your game with this subpar piece of software. pbs is such an integral part to safe operations, the current state is just not cutting it.