does PBS need to be into a cluster
- Thread starter Indirectelex
- Start date
Hey,
no, there is no such thing as a PBS cluster. You can however sync backups between different PBS servers[1].
[1] https://pbs.proxmox.com/docs/managing-remotes.html
no, there is no such thing as a PBS cluster. You can however sync backups between different PBS servers[1].
[1] https://pbs.proxmox.com/docs/managing-remotes.html
Hi there,
I had PBS into a VM, but I had to upgrade proxmox 7.1 to 7.2. I had no choice but to download proxmox-7.2.iso and install it. This situation makes PBS into a VM useless. If I install PBS on another machine I guess I should create a cluster, right?
I had PBS into a VM, but I had to upgrade proxmox 7.1 to 7.2. I had no choice but to download proxmox-7.2.iso and install it. This situation makes PBS into a VM useless. If I install PBS on another machine I guess I should create a cluster, right?
If you want migrate the VM to the new machine, then yes, those two machines would have to be in a cluster. However, you can also create a new PBS VM on the new machine and sync[1] all the backups to the new PBS and delete the old one, which is IMHO the better option, especially if you'd create the cluster only to move that one VM.
[1] https://pbs.proxmox.com/docs/managing-remotes.html
[1] https://pbs.proxmox.com/docs/managing-remotes.html
I got a problem with the clusters because I can't access one machine if the other is not powered up, does it still that way with pve 7.2?
Yes, the cluster needs to be quorate all the time [1].
If you, for example, only have two nodes in your cluster, it is highly recommended to setup a QDevice [2].
[1] https://pve.proxmox.com/wiki/Cluster_Manager#_quorum
[2] https://pve.proxmox.com/pve-docs/chapter-pvecm.html#_corosync_external_vote_support
Does that means one node will be able to be offline while I can use the other one online?
Yes, because the qdevice will act as a third voter. So if onesof two nodes is offline the remaining node still got 2 of 3 votes and quorum is given.
Without the qdevice there are only two voters so when one node goes down the remeining node only got 1 of 2 votes so there is no majority and the remaining node will stop working until there is quorum again.
Last edited:
Should I start the configuration of the qdevice before or after I have made a cluster?
Hm, from the quorum being in section HA and not in section Cluster of the webinterface I would have thought I would have to enable HA separately for the quorum to become necessary. Does that mean a cluster is always HA, there is no way to just have a loose association of nodes that I can move VMs between? Would
pvecm expected 1 help as an ugly workaround, or will that definitely lead to problems when both nodes are up and try to sync? Can I move VMs to a different pve by using the pbs as a stepstone?Background: I want to have a pve on the machine the backup server runs on so that, in case the main machine fails horribly, I can restore the important VMs from the pbs to that machine's pve and run them there until I get the main machine back online. (If the backup server fails, the main server should continue running undisturbed.)
Now the questions I have are:
- do I run the pbs directly on the pve server or in a vm?
- if in a vm, qemu/kvm or lxc? Can both see the LTO drive?
- cluster or not?
A PVE-cluster has to be quorate all the time, even without using HA.
Currently not, but it is on the roadmap.
I have no (long term) experience with it, but what I read (from the Proxmox-Staff) it is absolutely and highly not recommended at all and may lead to problems sooner or later.
Yes.
I personally run it directly aside PVE. Therefore no fiddling with the storage(s) for the PBS-datastore(s) and, in your case, also not for the LTO-drive.
But others may recommend different.
If you can not (or do not want to) guarantee, that your cluster is quorate all the time, I highly would not recommend running a cluster.
This: [1] recent thread might be of interest.
[1] https://forum.proxmox.com/threads/best-practices-in-network-configuration.117961