Node showing ?

J

Joshuam08

New Member
Dec 24, 2025
2
0
1
Hi all,

I have just setup a cluster with two nodes. If I login to my PVE01 node, both nodes show in the cluster with green ticks (ie working). However, if I login to the PVE02 node, PVE01 shows as a grey question mark (same with the LXC on that node), but PVE02 has a green tick.

I've tried restarting pvedaemon and pvestatd on both hosts with no change. When I am logged into PVE02, even though it shows as ?, I can still view LXC resource being used, access shell etc on that node.

Any ideas?

Thanks!
 
Just guessing: make sure that name resolution works correctly. Usually by a full blown DNS server (possibly overkill) or (simpler) by maintaining a consistent /etc/hosts, containing all host names and addresses of all hosts. That file should be identical on all hosts.

"ssh nameofpve1" + "ssh nameofpve2" must work on both nodes without asking for a password.

"pvestatd" is important. Check it on both machines with "systemctl status pvestatd.service". Already done...


Sidenote: a cluster with only two nodes will not work reliably. Search for "Quorum Device"...
 
UdoB said:
Just guessing: make sure that name resolution works correctly. Usually by a full blown DNS server (possibly overkill) or (simpler) by maintaining a consistent /etc/hosts, containing all host names and addresses of all hosts. That file should be identical on all hosts.

"ssh nameofpve1" + "ssh nameofpve2" must work on both nodes without asking for a password.

"pvestatd" is important. Check it on both machines with "systemctl status pvestatd.service". Already done...


Sidenote: a cluster with only two nodes will not work reliably. Search for "Quorum Device"...
Click to expand...
Thanks for the reply! I can confirm DNS is good. I have both a full DNS server setup but also already modified the host file with both nodes for good measure and the issue persists. I can ping node name and node.tld no problem from both hosts. I can also SSH from both nodes with no password required.

In terms of the side note, I only clustered the nodes so I could manage either node from one webpage. I do not plan on using any HA, fail over etc. Is the quorum device still required if I am not using HA functions?
 
Joshuam08 said:
I only clustered the nodes so I could manage either node from one webpage. I do not plan on using any HA, fail over etc. Is the quorum device still required if I am not using HA functions?
Click to expand...
Well..., yes.

As soon as one of two nodes is unavailable you cannot administrate the single survivor - it will deny everything with "no quorum". Just try it!

There are workarounds for disaster recovery: "pvecm expected 1" will probably work. But that's meant to be a workaround, not a solution! It can not guarantee the cluster-wide consistency as the state of the other (dead) node is officially "unknown". For that you need a third vote...
 
You must log in or register to reply here.
Top Bottom