Bareos Checkpointing feature merged - Impact on Spooling?

[Brock Palen]

I was testing some recent builds and I nocited

26-Sep 08:15 myth-sd JobId 66683: Doing timed backup checkpoint. Next checkpoint in 60 seconds
26-Sep 08:15 myth-sd JobId 66683: Checkpoint: Syncing current backup status to catalog
26-Sep 08:15 myth-sd JobId 66683: Checkpoint completed

I looked at the changelog and PR this looks great! One of the biggest issues using Bareos over slow connections was doing larger backups.

There are several questions,

1. will it let failed backups (max time limit window ) resume where it left off? If so is that only for Incremental what about Fulls?

2. Spooling, the log states checkpoints to catalog before any messages about spooling. How does it make sure that data in the catalog is committed to the volume and not lost in a spool eg. SD dies, or using large spools nothing is ever written to the volume before the backup fails?

There doesn’t appear to be anything in the draft documentation about how checkpoints interact with spools, or not having a completed “Full” backup.

Thanks! This is great work!


Brock Palen
bro...@mlds-networks.com
www.mlds-networks.com
Websites, Linux, Hosting, Joomla, Consulting

[Alaa Eddine Elamri]

Hey Brock!

Thank you for your feedback!

1- For the time being, checkpoints' only role is to save data to be able to restore failed or cancelled jobs, but do not resume them yet. It will also be used to monitor backup progress. Resuming failed jobs is indeed planned for the future

2- With spooling, checkpoints are not doing anything, the messages in the log say that a checkpoint is occurring, but deep down it is not. A new PR is on the way to make some needed documentation updates about checkpoints, the logging will be adjusted to show the right messages, and checkpoints will be just deactivated when using spooling for now.

If you have other questions or remarks while using checkpoints, let us know.

Regards,
Alaa Eddine.