[SOLVED] [vm backups] Where are the full backups of my vms?

[Sebi-S]

Hello everbody,

we are using pbs 2.2.5 and I have to admit that I don't fully understand the backup solution yet.

My question in a nutshell: "Where do I find the full backups of my vms?"

I know that pbs uses incremental backups, but as far as I have used incremental backups in the past, I have always had to work with a full backup and incremental backups based first on the full backup and then on the respective incremental backup.
On the pbs server I found the storage directory containing the vms' incremental backups. No problem so far.

My point is: using only incremental backups is fine, but having a corrupt incremental backup in a long chain of backups leads to a disaster. So I'd like to know how the backups actually work and how/if I can create full backups of the vms.

(I've read the manual and watched a few videos before opening the thread, but I didn't fully understand it yet)

 

[Dunuin]

Hello everbody,

we are using pbs 2.2.5 and I have to admit that I don't fully understand the backup solution yet.

My question in a nutshell: "Where do I find the full backups of my vms?"

I know that pbs uses incremental backups, but as far as I have used incremental backups in the past, I have always had to work with a full backup and incremental backups based first on the full backup and then on the respective incremental backup.
On the pbs server I found the storage directory containing the vms' incremental backups. No problem so far.

My point is: using only incremental backups is fine, but having a corrupt incremental backup in a long chain of backups leads to a disaster. So I'd like to know how the backups actually work and how/if I can create full backups of the vms.

(I've read the manual and watched a few videos before opening the thread, but I didn't fully understand it yet)

All backups are full backups. There are no differential backups. But the full backups are still incremental (atleast for VMs, only partially for LXCs s these can't use dirty-bitmapping) because of deduplication and dirty-bitmapping. And because of deduplication there is no single big image file. Everything is chopped into small chunks of max 4MB in size. Lets say you got a average chunk size of 2MB, then your full backup of a 2TB VM will result in an index file pointing to chunks and 1 million chunk files. Thats why you should use SSDs and not HDDs because that needs alot of IOPS performance. And because of deduplication all backups share the same chunk files, so there is no way to for example copy a single backup to another storage (except from doing a sync job between two PBS datastores).
You should do verify and reverify jobs. These will read the index file that is pointing to the chunks, will read the chunks and rehash them and compare the calculated hash to the stored hash. So this will find missing chunks or corrupted chunks.

 

[Sebi-S]

That makes sense. Thanks a lot!