Recent content by iwik

Yes, but during upgrade cycle you always have have some versions mixing. And sometimes it works, even if it is unsupported/untested. But not this case - and it is important to know how it behaves.

Well, they are older, like pve-manager/8.2.2 mostly. I can confirm issue is with versions mixing. Recycling name was not the issue. Upgraded one of older v8 to latest 9.2.10 and same problem. I started with upgrades, it should solve it.

Hi all, I added a new node to my existing PVE cluster ( mixed versions — most nodes are 8.x, one is still 7.x, and the new node is 9.2.2). Context: the node name "pve8" was previously used by a different physical node, which I properly removed from the cluster via `pvecm delnode` some time...

But technology used for this (implementation of this feature) are not lvm snapshots.

This really interesting. It seems it implements lvm as starlvm and enables snapshots on this lvm type. So something that proxmox does not want to do. https://repo.starwind.com/proxmox/pool/main/s/starwind-proxmox-plugin-pve9/ starwind-proxmox-plugin-pve9_1.1.36.deb 05-Mar-2026 16:20 8748 size...

PVE dev team knows about this, right? Will be there some change for default settings when QCOW on lvm is used?

Same here :) EDIT: Also, when you are using veeam, backup fails with general error Failed to connect the NBD server to the hypervisor host So let's it write here, so others can find it, since I did not google it anywhere, this can be root cause.

That is due backups are "pushed" into PBS, not scheduled and managed by PBS. PBS works just like storage space, it does not know if job was run or not.

Inspired by a forum solution for FreeIPA, I created patch for AD using AI Claude. Code is attached in this bug report.

So, now I know why GFS2 is unsupported from proxmox side... It works and performance was also very good, but it breaks after some time (more than half a year). Then it is pain, when everything is stuck and all pve nodes need to be restarted and then fsck run. This is what you don't want from VM...

Hi using gfs2 a today hit bug: [2479853.036266] ------------[ cut here ]------------ [2479853.036509] kernel BUG at fs/gfs2/inode.h:58! [2479853.036721] Oops: invalid opcode: 0000 [#1] PREEMPT SMP PTI need to restart pve node. :-/

Latest ceph release (20) removes both: https://ceph.io/en/news/blog/2025/v20-2-0-tentacle-released/#changes MGR Users now have the ability to force-disable always-on modules. The restful and zabbix modules (deprecated since 2020) have been officially removed.

When 3 node CEPH cluster is setup using full mesh - routed mode, is it possible to use this ceph network for migration as well? In GUI I have to select interface, but actually there are two interfaces in this case, each to different node.

I think it is fine, if you will not use zfs. Personaly I am using mdadm raid 1 (yes I know...), 2x MX500 2TB and LVM on top for VM. Running fine for 3 years.

You can do it by disk passthrough (cli). https://pve.proxmox.com/wiki/Passthrough_Physical_Disk_to_Virtual_Machine_(VM). Since FC SAN devices usually have multiple paths, you have to configure multi path https://pve.proxmox.com/wiki/Multipath#Set_up_multipath-tools and then pass to vm this dm...