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 ago. I've now reused that name for a new node, installed fresh with PVE 9.2.2, with a new/different IP address.
The join itself completed cleanly via `pvecm add`:
waiting for quorum...OK
(re)generate node files
generate new node certificate
merge authorized SSH keys
generated new node certificate, restart pveproxy and pvedaemon services
successfully added node 'pve8' to cluster.
TASK OK
After joining, cluster status looks healthy from every node I check:
- `pvecm status` — quorate, correct number of votes
- `corosync-cfgtool -s` — no faults on links
- `/etc/pve/.members` — new node listed with `"online": 1` and correct (new) IP, consistent across all nodes I checked
However, in the web GUI, the new node shows as offline/grey/"unknown" when viewed from the older nodes (the 7.x node, and the 8.x nodes). Viewed from the new node itself (9.2.2), everything looks fine — online, normal CPU/RAM/uptime stats.
So corosync/pmxcfs membership looks fully healthy, but something above that layer isn't propagating/rendering correctly to the older nodes' GUI.
Already checked/ruled out:
- Quorum/corosync connectivity — fine (see above)
- Time sync between nodes — no significant drift
- `pvestatd` on the new node — active, no errors in journal
Questions:
- Is this a known limitation when an older PVE GUI (7.x/8.x) tries to display a node running a newer major version (9.x) — i.e. it can't parse the newer status broadcast format?
- Could reusing a previously-removed node name (with a new IP) leave any stale reference behind that would cause this specifically, versus just adding a brand-new node name?
Thanks for any input!
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 ago. I've now reused that name for a new node, installed fresh with PVE 9.2.2, with a new/different IP address.
The join itself completed cleanly via `pvecm add`:
waiting for quorum...OK
(re)generate node files
generate new node certificate
merge authorized SSH keys
generated new node certificate, restart pveproxy and pvedaemon services
successfully added node 'pve8' to cluster.
TASK OK
After joining, cluster status looks healthy from every node I check:
- `pvecm status` — quorate, correct number of votes
- `corosync-cfgtool -s` — no faults on links
- `/etc/pve/.members` — new node listed with `"online": 1` and correct (new) IP, consistent across all nodes I checked
However, in the web GUI, the new node shows as offline/grey/"unknown" when viewed from the older nodes (the 7.x node, and the 8.x nodes). Viewed from the new node itself (9.2.2), everything looks fine — online, normal CPU/RAM/uptime stats.
So corosync/pmxcfs membership looks fully healthy, but something above that layer isn't propagating/rendering correctly to the older nodes' GUI.
Already checked/ruled out:
- Quorum/corosync connectivity — fine (see above)
- Time sync between nodes — no significant drift
- `pvestatd` on the new node — active, no errors in journal
Questions:
- Is this a known limitation when an older PVE GUI (7.x/8.x) tries to display a node running a newer major version (9.x) — i.e. it can't parse the newer status broadcast format?
- Could reusing a previously-removed node name (with a new IP) leave any stale reference behind that would cause this specifically, versus just adding a brand-new node name?
Thanks for any input!
