Cluster Configuration Failed? "communication failure (0)"

[limone]

Hi,

I have two nodes which I want to manage via one interface, therefore I created a cluster.
Because the two hosts are not in the same network I configured an openvpn connection between them.

Node1 got the IP 10.0.205.1
Node2 got the IP 10.0.205.251

On node one I did this:

pvecm create vpncluster -bindnet0_addr 10.0.205.1 -ring0_addr alpha

On node two I did this:

pvecm add alpha -ring0_addr bravo

I added these hosts in both nodes /etc/hosts file

#corosync
10.0.205.1      alpha
10.0.205.251    bravo

So far so good, everything added just fine.

/etc/pve/corosync.conf:

Spoiler

logging {
  debug: off
  to_syslog: yes
}

nodelist {
  node {
    name: hostnameofnode2
    nodeid: 2
    quorum_votes: 1
    ring0_addr: bravo
  }

  node {
    name: hostnameofnode1
    nodeid: 1
    quorum_votes: 1
    ring0_addr: alpha
  }

}

quorum {
  provider: corosync_votequorum
}

totem {
  cluster_name: vpncluster
  config_version: 2
  ip_version: ipv4
  secauth: on
  version: 2
  interface {
    bindnetaddr: 10.0.205.1
    ringnumber: 0
  }

}

The corosync.conf is getting synced between the nodes, right?
Because when I open it on node one and want to open it on node two I get a warning message it's already opened.

But now my actual Problem: In the Webinterface of node1 I can see node2, but I cannot acces it. No information is getting displayed and I am not able to create a VM.
Given errors: "communication failure (0)" or "Connection timed out (595)"
From the webinterface of node2, however, I can manage node1 completely.

Is there any way to spot the problem?
I have to manage from webinterface of node1, if I could use node2's webinterface I wouldn't create this thread

 

[wolfgang]

Hi,

First this is not supported and will probably never proper work over openvpn.

1.) ensure the latency is under 2ms.
2.) use unicast instead of multicast see man corosync.conf

 

[limone]

Thanks for your reply.

1.) Ping is 30ms, the servers are very far away from each other.
2.) unicast is not helping in this situation, stillt the same error

 

[limone]

But very strange that everything works flawless from the second nodes webinterface.. Maybe it has something to do with hostname lookups?