PM 1.8, Clustering (PVECA) not working on remote servers

[mylesw]

I have four (4) PM 1.8 servers all running identical PM versions. Two are located on the same LAN and 2 are on remote servers. No issues with joining the two serves on the LAN, but I'm not having any luck with the remote servers.

Thinking its a firewall issue, I've opened port 22 on both the master & slave sides, and tested through firewalls to make sure its all working. I can telnet from one to the other on port 22 and its communicating just fine.

But when I attempt to add the secondary slave server to the master, with:

pveca -a -h xx.xx.xx.xx

I get back:
host 'xx.xx.xx.xx' is not cluster master

I check on the master, however and its showing the error from the attempted connection as:

ERROR: 500 Can't connect to 127.0.0.1:50001 (connect: Connection refused)

What is going on here? Are there are additional ports that need to be opened on firewalls other than 22?

Myles

 

[mylesw]

One further piece of information.... The secondary server that I'm attempting to connect to the master is in a different timezone to the master. This server is in US Central Std Time and the master is in US Arizona (MST). There is a 1 hour time difference between both servers, however both servers have their time clocks synch'd through NTP.

 

[mylesw]

More info... The fourth PM server which is located in the same data center, but is not behind the same firewall as the two servers that have been connected together in cluster and are working great, and is in the same timezone as the master, also fails with the identical error as the remote server. So I believe this is some form of firewall issue. Not sure what the ERROR: 500 Can't connect to 127.0.0.1:50001 (connect: Connection refused) is referring to.

 

[dietmar]

Not sure what the ERROR: 500 Can't connect to 127.0.0.1:50001 (connect: Connection refused) is referring to.

Is pvetunnel service running on all nodes? Try

# /etc/init.d/pvetunnel restart

 

[mylesw]

Is pvetunnel service running on all nodes? Try

# /etc/init.d/pvetunnel restart

Interesting... I checked it on the master, and its running (shows with a ps ax | grep pvetunnel). That has one slave machine connected to the cluster successfully, so I think the master is working just fine.

Checked on the remote host and its not running there. I restarted it, but its still not running. I'm assuming it should be running all the time and showing in the process list. What is showing on the remote machine in syslog is:

Jun 12 08:39:26 proxmox3 pvetunnel[6424]: not starting server - not part of cluster

I suspect this is because it didn't succeed with adding the server to the cluster before.

Myles