PBS setup Best practices Questions

[Glowsome]

So after having met with PBS and wildly excited about it i have a question as to "best setup"

At current i have One PBS box, but have two separate PVE installations, one remote, one local.
I have separated the backups with NameSpaces, so there is no collision possible regarding backup-operations.

Now what i am confused about is regarding "best practice":

- Is it worth installing/configuring a separate PBS-box on the remote, let the backups be made locally and then configure a remote on the PBS locally ?
- Pro's/Cons ?
- If separate install remote PBS is recommended then as LXC, VM or natively on the PVE box ? ( local i installed PBS as LXC with pass-through to local storage)

Glowsome

 

[Dunuin]

- Is it worth installing/configuring a separate PBS-box on the remote, let the backups be made locally and then configure a remote on the PBS locally ?
- Pro's/Cons ?

Yes, and I would recommend to let both PBS sync each other. That way you can implement ransomware protection (see: https://pbs.proxmox.com/docs/storage.html#ransomware-protection-recovery), backups/restores will be faster, fast enough for live restores and in case a chunk corrupts you can resync the backup snapshots to fix it. Always good to have a second copy...especially offsite. And keep in mind that deduplication means no chunk of a datastore will be stored twice. So losing a single 4MB chunk could mean hundreds of backups taken over many years could be broken as all share the same single chunk. So good to not rely on a single datastore/PBS.

- If separate install remote PBS is recommended then as LXC, VM or natively on the PVE box ? ( local i installed PBS as LXC with pass-through to local storage)

LXCs aren't tested internally and therefore not recommended. Could break every update. I prefer a VM but bare metal next to PVE would also work.

 

[Glowsome]

LXCs aren't tested internally and therefore not recommended. Could break every update. I prefer a VM but bare metal next to PVE would also work.

Bare Metal is not an option, IMHO as that would only introduce a SPOF.

Now to be exact in my experience, i have done pass-through on LXC's regarding storage, but never on a VM, as the underlying storage is there, and want to utilize it as such. (we are talking about a Cluster here, tested fail-overs of the LXC with retaining access to the underlying (shared) storage)
For Local i had chosen (as experiment) LXC for PBS, with having a cluster-storage resource made available to it, and its been working out quite good to be honest.

I like to test around possibilities, so am also willing to take a chance, hence the choice for going the LXC-way.

For the remote at the moment i have 2 backup-jobs, one utilizing the native PVE backup (so full) for 3 days backing to local (remote)storage, and then a PBS backup to the Local for 7 days.
So in essence i have the 'local backup + having an off-site backup in place.

- Glowsome

 

[Glowsome]

So to come back to how i've set it up now with the suggestions being made (not honoring all):

- Both PVE instances will backup each day (3 day history) in a traditional way.
- Both PVE instances will backup each day (7 day history) via Local PBS on their local backup store.
- PBS remote will be synchronized to PBS local once a day after local PBS backups were done.

As you can see i have taken the (even tho not advised) way of going for a LXC to install PBS in, and within that using passthrough to the local storage.
This seemed for me the most logical way to not have more then one different implementation of a/the product.
As fallback therefore traditional backups still exist, just with a shorter retention-period.