Hi,
This is not a PBS question, which is why it's not in the PBS section of the forum.
Following a few bad design decisions, and correction to those decisions, I am left all my VM guest on a shared NFS storage (which is fine).
The problem is the shared storage is a manually mounted NFS on a PBS. Yes, you read that right. The PBS has been "decommissioned" so far as it's still installed, but really is at this point just a Debian install that happens to have PBS installed on it. PBS itself is not used for anything backup related. I`ll call it my "Frankenstein" system for clarity (to differenciate it from my real, working PBS system).
I feel it would be more useful to have that system setup as a SAN-type of thing, like TrueNAS (TrueNAS Scale, probably). I do have some basic experience with TrueNAS Core and ZFS.
I would like to, without having to move data to another server (I do not have one with the required space available), transform this PBS into a TrueNAS, and then reconnect the VM guests to this shared storage. Luckily (I think) I am using ZFS pools.
The specifics:
- The Frankenstein server runs PBS (as previously stated) which I do not need anymore since I have another instance running and setup properly.
- The boot drives are SSDs, in a ZFS mirrored pool
- The data drives hosting the NFS storage are a bunch of Mirrored VDEVs (RAID10-style) in a separate ZFS pool. There is no data except VM guest hard drives on this pool.
- I do have a backup of all the VMs on another well-setup PBS, but I would rather not have to restore them (more time consuming than just a ZFS export/import). Whiler I can handle some downtime, I would rather it be measured in minutes than hours.
- My hosts are on pve 8.1.3
My plan, which I would love constructive criticism on, is this:
1) Turn off all VMs on the PVE cluster (I can handle some downtime late at night)
2) Turn off the PVE hosts (needed?)
3) Export the VM ZFS pool on the Frankenstein system
4) Reinstall TrueNAS over the Frankenstein boot drives, turning this into a TrueNAS system
5) Import the ZFS pool in TrueNAS,
6) Setup NFS in TrueNAS , making the VM hard drives now reachable and available for use
7) Turn on cluster and somehow (?) tell the cluster that the VM hard drives are now to be found on the TrueNAS system (different IP? Different mount?).
8) Turn on the VMs
My main questions are :
- Step 7 - I actually would appreciate some idea on how to do that. I have no clue how to tell an existing VM "hey, sorry buddy your disks are now there". I can set it up with the previous PBS IP address I guess, but I'm not sure I can mount the drives in the exactly same place. I'm less confident here than the other steps.
- Do I need to turn off the PVE hosts in step 2? Or can they be left one, will notice the storage is gone, then suddenly see it come back when I reach step 6?
Of course any other useful comment is welcomed.
This is not a PBS question, which is why it's not in the PBS section of the forum.
Following a few bad design decisions, and correction to those decisions, I am left all my VM guest on a shared NFS storage (which is fine).
The problem is the shared storage is a manually mounted NFS on a PBS. Yes, you read that right. The PBS has been "decommissioned" so far as it's still installed, but really is at this point just a Debian install that happens to have PBS installed on it. PBS itself is not used for anything backup related. I`ll call it my "Frankenstein" system for clarity (to differenciate it from my real, working PBS system).
I feel it would be more useful to have that system setup as a SAN-type of thing, like TrueNAS (TrueNAS Scale, probably). I do have some basic experience with TrueNAS Core and ZFS.
I would like to, without having to move data to another server (I do not have one with the required space available), transform this PBS into a TrueNAS, and then reconnect the VM guests to this shared storage. Luckily (I think) I am using ZFS pools.
The specifics:
- The Frankenstein server runs PBS (as previously stated) which I do not need anymore since I have another instance running and setup properly.
- The boot drives are SSDs, in a ZFS mirrored pool
- The data drives hosting the NFS storage are a bunch of Mirrored VDEVs (RAID10-style) in a separate ZFS pool. There is no data except VM guest hard drives on this pool.
- I do have a backup of all the VMs on another well-setup PBS, but I would rather not have to restore them (more time consuming than just a ZFS export/import). Whiler I can handle some downtime, I would rather it be measured in minutes than hours.
- My hosts are on pve 8.1.3
My plan, which I would love constructive criticism on, is this:
1) Turn off all VMs on the PVE cluster (I can handle some downtime late at night)
2) Turn off the PVE hosts (needed?)
3) Export the VM ZFS pool on the Frankenstein system
4) Reinstall TrueNAS over the Frankenstein boot drives, turning this into a TrueNAS system
5) Import the ZFS pool in TrueNAS,
6) Setup NFS in TrueNAS , making the VM hard drives now reachable and available for use
7) Turn on cluster and somehow (?) tell the cluster that the VM hard drives are now to be found on the TrueNAS system (different IP? Different mount?).
8) Turn on the VMs
My main questions are :
- Step 7 - I actually would appreciate some idea on how to do that. I have no clue how to tell an existing VM "hey, sorry buddy your disks are now there". I can set it up with the previous PBS IP address I guess, but I'm not sure I can mount the drives in the exactly same place. I'm less confident here than the other steps.
- Do I need to turn off the PVE hosts in step 2? Or can they be left one, will notice the storage is gone, then suddenly see it come back when I reach step 6?
Of course any other useful comment is welcomed.
Last edited: