bacula.md

Backup system.

Contains 5 components all of which may optionally reside on separate machines:

1. "file server/daemon": an app that reads files from data source and transfers to a storage server.
2. "storage server", storage: reads data from a file server and writes to the backup storage (disk, tape, etc).
3. "database server", catalog: a catalogue of backups/files on the storage server.
4. "backup server", director: an app controlling services operation and running schedules.
5. Tray monitor and console: administration clients.

Configuration

Sections

Sections may refer to each another, which is used for config deduplication.

bacula-dir.conf

• Director
• Job: a backup, restore or other operation for Bacula to perform. May refer to FileSet, JobDefs, Pool. It may not write to a volume but to a pool.
• JobDefs: "job defaults", a template for Job. May refer to Schedule.
• FileSet: files to back up. Referred to by Job
• Schedule: time when job gets run.
• Client: connection config for a "file service".
• Autochanger: Idk exactly, but it's something about grouping multiple storage devices into a single unit.
• Catalog: connection config for a "database service".
• Messages: logging: things to go to email and to log, separately.
• Console: console configs, may be restricted in different ways for e.g. just status monitoring.
• Pool: a group of "volumes".
  • A volume is a storage device like a disk or a tape and it may not be referred directly by a job.

Misc

• "Restoring": done with restore command which uses "jobid" of the jobs log to restore from. But for some reason it also requires a Job with Type = Restore to be present, otherwise restore won't work. In particular, the FileSet directive is completely ignored (but required to be present), and instead you'd have to manually chose which files you want to restore.

Gotchas

• Don't confuse jobs and jobs! Already confused, are you? In Bacula the top-level thing represented by Job { config paragraph is called a job. And then if you ran a job multiple times, its history entries are called jobs as well! So when you see word "job" it may mean any of those concepts.

• Increasing a pool size won't propagate to volumes till you exec update volume in console.

• If you don't start postgresql before installing Bacula, then initial installation will get completely screwed up on the database side. Bacula says that to fix this you'd have to run some dbconfig-common after starting postgresql service, but Idk where this script (or even if it's a single script or a group).

  So happened that I didn't know any of that, so below is my unfinished research on how to fix this. The errors below are fixed in sudo -u postgres psql. You'll be getting:

  1. Unable to connect to PostgreSQL server. This is because it didn't create a user that's written in the config. Fix with create user bacula password 'insert_pswd';
  2. Unable to connect to PostgreSQL server. This time it's because it didn't create a database, fix with create DATABASE bacula;

  There will be more errors, but at that point I decided to completely remove the Podman/Docker image with bacula to get it reinstalled. Worth noting that simply apt purge bacula and reinstalling it won't be enough, it leaves some configs at random places that you'd have to search for. Yeah, that sucks.