proxmox datastore restore/rollback question

[RolandK]

let's assume we have a proxmox backup server where the datastore is getting it's own backup or snapshot at volume / filesystem level.

for example pbs datastore is on zfs and getting snapshotted.

now, what will happen if that datastore is getting a restore/rollback or recovery from backup, i.e. if being moved back in time ?

1. i guess proxmox-backup-server will need a restart aftwards?

2. what will happen when the next backup cycle starts? will dirty bitmaps getting invalidated and a full backup being forced ?

 

[Dunuin]

3.) what will happen if a snapshot of a datastore gets rolled back while a GC is running? Lets say phase 1 marked some chunks, then the chunks aren't marked anymore because of the rollback and then phase 2 starts deleting unmarked chunks.

 

[RolandK]

do we get corruption this way or not ?

is a subsequent incremental backup run sane ?

i guess they aren't , because i cannot see that the backup client has knowledge what was the last backup snapshot and without that, it assumes the backup server has the last valid snapshot where the VMs block change bitmap corresponds to.

 

[RolandK]

>i guess they aren't , because i cannot see that the backup client has knowledge what was the last backup snapshot

apparently the client MUST have knowledge of that.

if i delete the last backup from the pbs, on next backup run the bitmap is being invalidated/cleared and a new full backup being done. good !

wouldn't it be helpful to show some info in the log to what snapshot a new dirty bitmap refers to ?