Pull replication snapshot management

TrueNAS General
, ,
21

From over 175k this morning to this:

root@truenas[~]# zfs list -t snapshot | grep tank | wc -l
4725

I’ll need to monitor over the next few days, but looks like a massive improvement.

22

Did you make it in addition to PBS or instead of? If the latter – why? I’m planning to set up PBS but still haven’t.

23

PBS doesn’t back up the Proxmox host (AFAICT), just the VMs on it. In fact, there’s almost no documentation on how to back up the host. This was my solution to do that.

1 Like
24

Agreed, I’ve once searched it myself. Seems like proxmox and truenas are considering boot-pool as something ephemeral.

And yet, plain zfs snapshots do not contain RAM (iirc snapshots from the proxmox ui do).
I’ve a suspicion that it can cause some problems for restoring machines which have been running during the snapshots. IMO, all these snapshots are similar to restoring from a power-loss moment. Did you have any issues so far?

25

…which makes sense for TrueNAS–download the config file, upload it to a new installation, and you’re good to go. There’s no equivalent mechanism for Proxmox, unfortunately.

Not as yet. But again, I’m not backing up VMs, just the host.

26

IIRC, each cluster node contains all the configs from the other node (doesn’t work for the manual host tweaks, though). But that won’t help in the case of natural DISASTAH or hacked cluster.

Ah, makes sense. I didn’t pay enough attention – you have clearly mentioned the OS pools. Thanks for the insight.

P.s. So how do you back up your VMs, then?

27

That’s PBS.

1 Like
28

Late to this party, but it may have removed all the snapshots that do not match the updated naming scheme.

29

@Dan, did you ever consider using the proxmox backup client for backing up the proxmox node itself? If you did, why did you choose zfsnap eventually?

30

I remember looking into backing the hosts up to PBS, but not (in any real detail) why I didn’t go that direction. Best I can remember, I couldn’t find much in the way of docs for that route.

31

@dan Thanks for addressing this scenario. I have known for some time I am soon to face the same scenario since I have a similar setup with a “pure” OpenZFS on Linux server getting pulled to my NAS. Snapshots are made and managed by sanoid there.

32

I don’t have PVE installed on ZFS (yet - PVE 9 might change this), so I use the proxmox-backup-client in the PVE host. Documentation can be found here:
Backup Client Usage — Proxmox Backup 4.0.11-2 documentation
The basic command is:

# proxmox-backup-client backup root.pxar:/ --repository backup-server:store1

But I include /etc/pve to capture configuration settings and such:

# proxmox-backup-client backup root.pxar:/ --include-dev /etc/pve --repository backup-server:store1

I haven’t automated this yet, so I just run it regularly from the CLI and enter the necessary password when prompted.

1 Like
33

I’ve read your guide once more. And I have 2 questions:

  1. How would you restore pve-node from this backup? It was asked in the comment of your wiki too.
  2. Have you ever tried to actually restore it?
34

I have not tried to restore it. If/when I need to, I’d expect zfs send tank/dataset@snapshot | ssh root@pve2 zfs recv or some variation thereof would do it.

1 Like
35

Perhaps you would also need to execute update-grub proxmox-boot-tool refresh.

1 Like