Homelab: What to do if a PBS host failed? What additional files to back up?

[EpicLPer]

Heya,

currently planning to move all my homelab servers over to Proxmox and I'm learning as I go, pretty new to PVE and PBS
A thought that crossed my mind is: What if a PBS host fails? I did search around the forum and found some scripts on GitHub and Co. for backing up configuration files from PBS/PVE, however these are almost all over 4-5 years old and I'm not sure if they apply anymore.
As this is a homelab environment I also plan to just install PBS on the PVE host directly to save myself a server or two and electricity cost that way, hence why I'm a bit more concerned about potential disaster scenarios and want to prepare for this even better.

My questions are:

• What happens if a PBS host fails? Can I still restore backups with a freshly set up PBS host or do I need more than just the raw backup files?
• If I need more than just the backups, what else is important to save somewhere else? (apart from fstab for my mounted NFS shares)
• How exactly can I do this? Are there any updated pre-made scripts (with monitoring/logging) built in?

Thanks already for your help!

 

[fabian]

the datastore itself is self-contained, you don't need anything else to make the contained backups accessible on a new PBS. it's a bit easier if you have the config files from /etc/proxmox-backup, since you don't need to re-configure the datastore and jobs and users/access rights.

if your backups are encrypted, you should also think about safe-keeping of the required key(s). the PBS admin guide has details for that:
https://pbs.proxmox.com/docs/backup-client.html#encryption

 

[Dunuin]

As this is a homelab environment I also plan to just install PBS on the PVE host directly to save myself a server or two and electricity cost that way, hence why I'm a bit more concerned about potential disaster scenarios and want to prepare for this even better.

You could use WoL/BMC in combination with VZDump hook scripts and the PBS API to boot up a dedicated Proxmox Backup Server when starting a backup task and shut it down afterwards. In that case electricity isn't the big problem anymore and a free/cheap 10+ years old office PC would do the job.

A thought that crossed my mind is: What if a PBS host fails?

For anything in production or even bigger homelabs you would run two PBS and sync them via their sync jobs to keep two copies of all backups. Best case one of those PBS is offsite. If then one PBS goes down or your screw things up, the other one would still be available and ready to restore from.
Keep in mind that nothing will be stored twice when only using a single PBS with a single datastore, because of deduplication. If you create 30 backup snapshots of the same VM and a single 4MB chunk file corrupts, you probably don`t lose a single backup snapshot but all 30 backups of that VM (and maybe even other VMs). Here it really helps to have additional VZDump Backups or a second datastore or even better another synced PBS.

 

[gfngfn256]

• What happens if a PBS host fails? Can I still restore backups with a freshly set up PBS host or do I need more than just the raw backup files?
• If I need more than just the backups, what else is important to save somewhere else? (apart from fstab for my mounted NFS shares)
• How exactly can I do this? Are there any updated pre-made scripts (with monitoring/logging) built in?

1. Yes, you can restore with just the raw backup file. See fabian's comment above.
2. If you're adamant on running PBS on the PVE host, (yes its more cost-efficient, not always backup-efficient), IMO you have no worries concerning PBS host failure - since you'll be running PBS as a VM on PVE. Just backup the working PBS VM directly to a different source (your NFS) - and you're good to go.

Your main concern maybe the PVE host itself (and you make some reference to this). What I do for this - from time to time (and always before a major host update), live-boot your host & dd the whole PVE boot drive zipped-image to another storage (again you could use your NFS). You'll have a bit of downtime - but in my opinion is well worth it in the long-run. Consider for this using a smaller boot drive, as it'll be quicker to image. (PS. I've never had luck with Clonezilla for this - it doesn't like the PVE boot/system/partitions mechanisms - so just use good old dd, I've restored many a PVE with it).

Your NFS is probably also another point of failure to consider. I probably don't need to tell you this. So further copies of any system-critical backups should also be kept, preferably at another site.