Error backuping 2 pve servers on 1 pbs and clarification of source in backup datastore

[antipiot]

Hello!

I did some test backuping 2 different PVE servers on 1 datastore and ran into some problems

my test lab looks like this:
pve1 -> backup -> pbs1 -> replication -> pbs2
pve2 -> backup -> pbs2

1st thing: running the backup task on pve2 to pbs2 land me to an error: " INFO: Error: backup owner check failed (root@pam != backup@pam) "

INFO: Error: backup owner check failed (root@pam != backup@pam)
ERROR: Backup of VM 999 failed - command '/usr/bin/proxmox-backup-client backup --repository root@pam@10.0.0.43:datastore2 --backup-type vm --backup-id 999 --backup-time 1596612774 qemu-server.conf:/var/tmp/vzdumptmp11655/qemu-server.conf' failed: exit code 255

2nd things, i feels all the backups get mixed, it's especially hard and confusing if the VM's have the same VMID:
Is there a chance to have a way to differentiate them? Either by an automatic tagging (eg: using hostname as prefix or something) or by using a tag slot?
borg-prune allow "--prefix='{hostname}-'" to be used, it may be a great thing!

I'm not used to the mechnasims implied but i feel there should be no error doing so.

Thanks for your work!

Regards

JS

 

[oversite]

Hello!

I did some test backuping 2 different PVE servers on 1 datastore and ran into some problems

my test lab looks like this:
pve1 -> backup -> pbs1 -> replication -> pbs2
pve2 -> backup -> pbs2

1st thing: running the backup task on pve2 to pbs2 land me to an error: " INFO: Error: backup owner check failed (root@pam != backup@pam) "

INFO: Error: backup owner check failed (root@pam != backup@pam)
ERROR: Backup of VM 999 failed - command '/usr/bin/proxmox-backup-client backup --repository root@pam@10.0.0.43:datastore2 --backup-type vm --backup-id 999 --backup-time 1596612774 qemu-server.conf:/var/tmp/vzdumptmp11655/qemu-server.conf' failed: exit code 255

View attachment 19004


2nd things, i feels all the backups get mixed, it's especially hard and confusing if the VM's have the same VMID:
Is there a chance to have a way to differentiate them? Either by an automatic tagging (eg: using hostname as prefix or something) or by using a tag slot?
borg-prune allow "--prefix='{hostname}-'" to be used, it may be a great thing!

I'm not used to the mechnasims implied but i feel there should be no error doing so.

Thanks for your work!

Regards

JS

It makes much more sense for me to have a separate datastore for each proxmox environment. For me the datastore is like a folder, it would confuse me a lot to mix VMs from two PVEs in the same folder, not the least when restoring an environment. I feel it's a simple enough solution to have another datastore for another environment.

 

[antipiot]

It makes much more sense for me to have a separate datastore for each proxmox environment. For me the datastore is like a folder, it would confuse me a lot to mix VMs from two PVEs in the same folder, not the least when restoring an environment. I feel it's a simple enough solution to have another datastore for another environment.

Well, i thinked about this but it's kind of hard to do so.
Imagine having 1 ZFS raid or 1 drive for every host you may want to backup or for every backup jobs.

Things would be much easier if there was a way to split / separate jobs in 1 datastore.

 

[oversite]

Well, i thinked about this but it's kind of hard to do so.
Imagine having 1 ZFS raid or 1 drive for every host you may want to backup or for every backup jobs.

Things would be much easier if there was a way to split / separate jobs in 1 datastore.

Oh noo, not one zfs pool for each, and also not for every host, just for every cluster. Dunno how many PVEs you have, but i have several hosts in one cluster, the cluster hosts are ofcourse sharing the VM-numbers.

However, what i ment was, as an example, create the zfs pool with the disks you want, like the pool will be /zfs1 on the pbs.
Now, create the datastores like /datastore/pve1 with path /zfs1/pve1 and another datastore /datastore/pve2 with path /zfs1/pve2. Now you have two different datastores on the same zfs pool. This way you cannot have any VM-numbers mixed while having all backups on the same zfs pool.
This is already dead simple for any combination, no need to complicate this.

 

[fabian]

Well, i thinked about this but it's kind of hard to do so.
Imagine having 1 ZFS raid or 1 drive for every host you may want to backup or for every backup jobs.

Things would be much easier if there was a way to split / separate jobs in 1 datastore.

datastores are just directories, you can have multiple datastores on a single ZFS dataset (or multiple datastores each on their own dataset on a single zpool) or single mountpoin.

 

[antipiot]

Oh noo, not one zfs pool for each, and also not for every host, just for every cluster. Dunno how many PVEs you have, but i have several hosts in one cluster, the cluster hosts are ofcourse sharing the VM-numbers.

However, what i ment was, as an example, create the zfs pool with the disks you want, like the pool will be /zfs1 on the pbs.
Now, create the datastores like /datastore/pve1 with path /zfs1/pve1 and another datastore /datastore/pve2 with path /zfs1/pve2. Now you have two different datastores on the same zfs pool. This way you cannot have any VM-numbers mixed while having all backups on the same zfs pool.
This is already dead simple for any combination, no need to complicate this.

datastores are just directories, you can have multiple datastores on a single ZFS dataset (or multiple datastores each on their own dataset on a single zpool) or single mountpoin.

Thanks!
Totally makes sense - feeling dumb now