[SOLVED] Cant start PVE console session on one node but can on the other

[yswery]

Hi all

I have a weird issue that I have one node in our cluster which is refusing to initialise a API novnc websocket (the other nodes work fine)

In the logs I see the following:

2024-01-10T13:02:30.304869+13:00 cook pveproxy[2390675]: connect to 'localhost:5900' failed: Connection refused at /usr/share/perl5/PVE/APIServer/AnyEvent.pm line 551.
2024-01-10T13:02:35.303735+13:00 cook pveproxy[2390675]: worker exit
2024-01-10T13:02:35.404032+13:00 cook pveproxy[2390662]: worker 2390675 finished

I tried the basic restarting pve-cluster.service and pveproxy.service
Also I can confirm that localhost in the /etc/hosts file points to 127.0.0.1 (like in all the other functional nodes in the cluster)

Does anyone know where i can look to see why this one node is having these issues via the API (using the PVE webui works fine)

 

[yswery]

Found the issue/solution.

Since we have a few nodes in the cluster what I was doing was

making a request to https://node1:8006 (picked at random from the pool of nodes) to create the vncproxy on a URI like /nodes/NODE2/lxc/{vmid}/vncproxy

This means that the VNC started on NODE1 (I was expecting it to start on NODE2) while the vnc proxy session starts on NODE2 (as I expect
This is also why we saw this sometimes work and sometimes wont work.

Not sure if this is a bug or by design, but the fix was to simply execute the vncproxy method using the specific node I am later querying for the websocket and not mix/match