Proxmox Backup Server resume sync

[Jarvar]

Curious if Proxmox Backup Server Remote sync is terminated prematurely, when the task is initiated again, does it resume from where it left off? or is the whole process terminated and needs to go through the sync all over again?
For example, Site A sync to remote Site B gets terminated or aborted after several hours.
If I go to sync again in the future will it pickup from where it left off?
Thanks.

 

[shanreich]

It should resume from where it left off, since the Sync transfers the chunks that compose the whole backup. There might be a small overhead for recalculating the missing chunks again, but this should be negligible compared to the duration of the whole sync job.

 

[stephan_h]

Related to the initial question, I’ve got the issue that my internet provider is disconnecting my internet connection once every 24h. I am now trying to initially backup a large LXC of 2TB from PVE to remote PBS. Starts like a charme, but the whole transfer would take approx. 80 hours. Unfortunately the backup process stops with an error when the the disconnect happens. Once I restart the backup, it seems that no data is being transferred up to the point where the error occurred (see attached screenshot). So I guess that this is the "small overhead" mentioned in the comment above. But actually it takes as long as the data transfer before in my case.

So, at the moment I don’t see a way how I can sync 2TB with the burden of a disconnection every 24 hours. Is there a workaround that I am missing?

 

[UdoB]

I am not sure, but: does the syncing process stop because the pure fact of a short interception? Or does it stop because an IP address has changed due to the restart - and the current session has been invalidated?

If the reason is the changed address I would circumvent it by establishing a Wireguard Tunnel. Those addresses wouldn' t change and the session would survive a short (not sure about timeout limits...) interruption due to any reasons like a restarted DSL modem or other hickups "on the internet".

Just an idea...

 

[fabian]

@stephan_h the OP asked about sync, your question seems to be about the backup itself. these two tasks are completely different under the hood.

for a (first!) backup the problem is that a lot of the work has to be done again, even if the PBS server already has many/most of the chunks from a previous, aborted backup attempt..

this part always happens for LXC backups:
- the backup client has to read all the data/files
- this stream of data is split into chunks
- the chunks are hashed (+compressed + optionally encrypted)

if there was a previous successful backup snapshot, the following can be done
- chunks which are part of the previous backup snapshot are just registered with the server
- other chunks are uploaded to the server

if there was no previous snapshot:
- all chunks are uploaded to the server

the server then always:
- discards chunks which are already stored on the server

so as long as there is no backup snapshot, an LXC backup will always start over from the beginning, but the server will save some writes. once a successful backup has been made, uploading of chunks that are already backed up can also be skipped. reading on the client side is currently always done in full, no matter what.

you basically have two options
- setup a local PBS, do the backup there (uninterrupted hopefully), then sync to your actual PBS server, and hope that incremental backups in the future work directly to the external PBS server (or keep the local+sync setup if you have enough local space)
- setup aggressive exclusion to keep the backup size down, do a backup, relax the exclusion a bit, do another backup, rinse & repeat until all data is backed up (don't forgot to then prune the "reduced" backups so you don't mistake them for regular ones!)

 

[stephan_h]

@UdoB : Thanks for your hint. Hard to say, but I guess it’s the disconnect itself (log of my router says it takes around 8 to 10 seconds until reconnection with new IPv4).
@fabian : Sorry for the misunderstanding regarding sync & backups. Option 1 is what I already had in mind. I will give it a try. Thanks a lot for clarification.

 

[ITT]

Same here, i had to sync around 12TB (VM) to an external PBS via 100MBit-Link.
The Task (running over 2 Months!) seems never ending because of some interruptions.
So i think, PBS does not resume properly (after interruption)?

 

[fabian]

as long as you don't run GC on the target, syncing will definitely resume (it will re-download some of the metadata, but the chunks will be detected as already there and not downloaded again). that being said, at full line speed that would take almost two weeks, and that is a very optimistic estimate and doesn't include overhead, pauses because the sync was interrupted but not yet restarted, etc..

 

[ITT]

Thanks fabian, disabled GC and hopefully this sync finishes sometime.