Tar Backups timing out on too large part number

[adrw]

@girish Alternatively, can we add configurability of parallel copy threads similar to what was added to the rsync advanced configuration options? If we can increase parallel copying than I think this tar backup task will succeed.

[girish]

@adrw said in Tar Backups timing out on too large part number:

Task 1735 timed out

@adrw Something else is going on. The above error message suggests that the backup task took more than 12 hours. Is that the case? Was backup running for more than 12 hours?

[adrw]

@girish Yes, pretty sure it was.

[girish]

@adrw In that case, the real issue is https://git.cloudron.io/cloudron/box/-/blob/master/src/backups.js#L1235 . There is a timeout of 12 hours for backup tasks. Can you bump it to like 36 or something?

For copy concurrency, there is already a slider for that. Don't you see it under advanced? Note that where it is failing now is that it has to copy the parts of the same file (basically, the single file is so big that we have to split it up into many parts). I have to check if parallel multi-part copy is allowed. If it is, it's easy to do.

[girish]

@adrw It seems parallel multi-part copy will work per https://dzone.com/articles/amazon-s3-parallel-multipart . Looks like a good change to make.

[adrw]

@girish The only advanced features I see for the tar backups is the memory, none of the parallel copy or upload/download like I see when I'm in rsync mode.

[adrw]

@girish Thanks! That's what I was looking for, I'll give that a shot.

[girish]

@adrw Ah indeed. When doing tar backup, there is only one file to upload and copy. Nothing to do in parallel (apart from the multi-part copy which I think can be hardcoded).

[adrw]

@girish Could the copy that happens at the end of the task (I think from the snapshot folder to the timestamped one) be done in parallel? It seems to be done serially right now which contributes to the longer task time to some extent.

[adrw]

@girish Could this timeout be configurable in a future release?

[girish]

@adrw I made the timeout now 24 hours. The timeout is really just there to kill "stuck" backups. This can usually only happen when there is a bug in our code.

[adrw]

@girish Great! Thanks again for your help debugging this and adding more configuration. Huge help for larger backups like mine.