[SOLVED] Issue with a node on a different network

[raztor]

Hi everyone,

I have a 3-node Proxmox cluster. Two nodes are on my local LAN (192.168.100.x), and the third one is colocated remotely, connected through a TAP VPN bridge (layer 2) on a vlan without any rules. The VPN connection has a stable latency of less than 5ms.

All nodes can ping each other and I can access the Web GUI of the remote node from the others without issues. The corosync service is running on all three, and the IPs in corosync.conf are set to use the VPN interface (10.0.11.x range).

However, the remote node always shows as offline in the cluster (pvecm nodes doesn’t list it), even though it's up and running normally.

I’ve verified that the ring0_addr is set correctly to each IP of each node, MTU is consistent, time is synced, and multicast should be working (since it's layer 2). Still, corosync doesn’t establish links with the other nodes.

the weirdest thing is that i can access the node from the web ui on the other nodes but it still shows as disconnected (see images).

Any idea what could be causing this or how to debug further. Does it really need to be on the same vlan or can i fix the issue?

Thanks in advance!


nodelist {
node {
name: lhotse
nodeid: 3
quorum_votes: 1
ring0_addr: 10.0.11.200
}
node {
name: lithium
nodeid: 1
quorum_votes: 1
ring0_addr: 192.168.100.2
}
node {
name: sodium
nodeid: 2
quorum_votes: 1
ring0_addr: 192.168.100.4
}
}


I ended up migrating the cluster to the same vlan of the remote node to avoid issues and it worked

 

[raztor]

i forgot to mention that if i enter the webui from the co-located node it's exactly the same but with the other 2 nodes disconnected

 

[UdoB]

Just guessing: make sure that hostname resolution works correctly. Usually by maintaining a consistent /etc/hosts, containing your tunnel addresses of the named hosts.

"ssh nameofpve1" + "ssh nameofpve2" + "ssh nameofpve3" must work on all three nodes without asking for a password.

 

[bl1mp]

Hi,
can you provide the result of

pvecm status

journalctl -xeu corosync

as well as the network configuration.

cat /etc/network/interfaces

of the three nodes?

BG