Schedule DAILY restore from PBS possible to standby mode? - need advices.. !

[DC-CA1]

hi we try to find a way to not lose the benefit from each of our proxmox cluster that use direct attached shared SAS . the respond time is ultra fast and we dont consume our Network interface for storage transit.

we use GFS2 Shared and we Can than do Snapshot for Qcow2 and Raw without any issue.

BUT ! in case of solid hardware outage etc , we would like to have the VM backup ready on a NFS array for example . as restoring VM from PBS will be too long .
i have been reading for days and i dont find a solutoin. i dont want to go ZFS local i it would cost so many $ in lost storage , and i cant do Snapshot of GFS2 system if it mounted on more than one node.

so i came with that idea this morning. having PBS restoring only the disk on a NFS target on a daily base would be a nice idea !

or do Rsync support Snapshot from active VM ?

anyway any idea would be appreciated
if nothing is possible we will end up starting considering CEPH but we cant believe there no solution to keep working in Direct attached ...

 

[fiona]

Hi,

hi we try to find a way to not lose the benefit from each of our proxmox cluster that use direct attached shared SAS . the respond time is ultra fast and we dont consume our Network interface for storage transit.

we use GFS2 Shared and we Can than do Snapshot for Qcow2 and Raw without any issue.

BUT ! in case of solid hardware outage etc , we would like to have the VM backup ready on a NFS array for example . as restoring VM from PBS will be too long .

i have been reading for days and i dont find a solutoin. i dont want to go ZFS local i it would cost so many $ in lost storage , and i cant do Snapshot of GFS2 system if it mounted on more than one node.

so i came with that idea this morning. having PBS restoring only the disk on a NFS target on a daily base would be a nice idea !

I mean, there is no integrated solution, but you can just query the datastore for the latest backups and restore them with a bit of scripting (e.g. using proxmox-backup-client list to get the latest backups and information about included disks and proxmox-backup-client restore to restore the disks). Then when you want to restore, you'll need to manually adapt the configuration file to reference those disks.

or do Rsync support Snapshot from active VM ?

anyway any idea would be appreciated
if nothing is possible we will end up starting considering CEPH but we cant believe there no solution to keep working in Direct attached ...

 

[DC-CA1]

@fiona thx for your return,

let say a fresh copy as been restored from a 500gb VM and is now sleeping, can the nexts daily scripted restore , reupload the vm from scratch or we may be able to manage to use the incremental backups instead ?

can we pay for that kind of script your profesionnals service ?

it would be a nice feature addon to be honest !

 

[fiona]

@fiona thx for your return,

let say a fresh copy as been restored from a 500gb VM and is now sleeping, can the nexts daily scripted restore , reupload the vm from scratch or we may be able to manage to use the incremental backups instead ?

Restore cannot be done incrementally.

can we pay for that kind of script your profesionnals service ?

Sorry, but no. I've got enough work already

it would be a nice feature addon to be honest !

I don't think there have been many requests in that direction and it's a rather special use-case. Many people use ZFS replication or Ceph to have better availability (of course backups should still be used additionally).

 

[DC-CA1]

The issue with zfs Replication is when you have more than 2 - 3 node it get ridiculus to consume that many lost storage ( 1:3 )

With ceph we lose the benefit of having vm accessing the data trough shared SAS connectivity.

So far the best way we found is doing shared sas with GFS2 . But we now want replicate the vm disks on another storage.

Is rsync able to snapshot or do incremental sync ?

Or maybe a Vss on top of that file array ?

 

[fiona]

The issue with zfs Replication is when you have more than 2 - 3 node it get ridiculus to consume that many lost storage ( 1:3 )

With ceph we lose the benefit of having vm accessing the data trough shared SAS connectivity.

So far the best way we found is doing shared sas with GFS2 . But we now want replicate the vm disks on another storage.

Is rsync able to snapshot or do incremental sync ?

I'm not an expert with rsync, but AFAIK it is made for file operations, not for dealing with (virtual) block devices, so I'd be surprised if it had such features.

Another example, what you could do: for qcow2, you could create a snapshot of the VM with qm snapshot (like that you get better consistency because of filesystem freeze if you have the QEMU guest agent configured and running in the VM) and then copy the snapshots of the disks with qemu-img dd -l <snapshot name> .... For raw, shut down the VM and copy with qemu-img dd.

Or maybe a Vss on top of that file array ?

Sorry, I never used that, so can't answer.

 

[DC-CA1]

we are working with Filelevel actually so your idea of Snapshot is not bad at all.

so we can Daily Rsync and keep a a 7 days SNAP on a NFS target + the full image ?

is there a way to schedule automated snapshot ?

 

[RobFantini]

we are working with Filelevel actually so your idea of Snapshot is not bad at all.

so we can Daily Rsync and keep a a 7 days SNAP on a NFS target + the full image ?

is there a way to schedule automated snapshot ?

have you looked at rsnapshot ?

it is good for data snapshots, like hourly, daily , weekly etc . I am not sure if it will do what you want for system.

 

[fiona]

we are working with Filelevel actually so your idea of Snapshot is not bad at all.

so we can Daily Rsync and keep a a 7 days SNAP on a NFS target + the full image ?

My suggestion with snapshots is completely orthogonal to rsync/file level. The VM's disk images might be saved as files, but from all I know, rsync can't look inside to see what's changed...

Or do you mean you want to just make copies of the filesystems within the disks rather than the full disk? In that case, rsync can be an option.

is there a way to schedule automated snapshot ?

You can use a cron job or set up a systemd timer.