(Configure) Retries during backup?

[philippms]

Currently I am backing up not only VMs from our Proxmox Infrastructure but also directories from our server to write them to tape in a second step.

These backups are around 100TB with big files in TB size, but some parts contain millions small files. So a typical backup job, even if partitioned in several directories with 10TB each, is taking days.

Sometimes, due to updates, very short network outtakes and so, a backup job breaks with

HTTP/2.0 connection failed
unclosed encoder dropped
closed encoder dropped with state
unfinished encoder state dropped
finished encoder state with errors
catalog upload error - channel closed
Error: connection reset

This means, we can retry the whole backup. Without any saved checkpoints or so.

Is there - maybe a beta configuration option - to enable a number of retries for - lets say - 10 minutes for failed chunks before breaking up?
Or is it possible to save an index file every 1h or so to be able to resume such a backup?

 

[Chris]

Hi,

Is there - maybe a beta configuration option - to enable a number of retries for - lets say - 10 minutes for failed chunks before breaking up?
Or is it possible to save an index file every 1h or so to be able to resume such a backup?

there is currently no such retry or resume functionality, the resume is tracked in: https://bugzilla.proxmox.com/show_bug.cgi?id=7051

These backups are around 100TB with big files in TB size, but some parts contain millions small files. So a typical backup job, even if partitioned in several directories with 10TB each, is taking days.

You can use change detection mode metadata [0], in an attempt to reduce overall backup runtime by only requiring to backup changed files again.

[0] https://pbs.proxmox.com/docs/backup-client.html#change-detection-mode

 

[philippms]

Yeah, Change Detection may work if a previous backup exist - but in our scenario there is no such previous backup, we are transferring data just to archive them to tape.

Thank you for the bugzilla link - such a checkpoint feature would be like "do a backup of the folder one by one with change detection metadata", I think this would rapidly speed up our archive backups if they fail ;-).