I'm trying to configure the firewall on the Proxmox nodes, but I'm definitely running into issues with the cluster.
Initially, after rebooting the nodes, everything works and looks fine. After about 10 minutes (sometimes more), it seems to stop working until I disable the firewall (then it shows green after a few seconds to a few minutes).
I have every node configured to accept ALL traffic from all the other nodes. It looks a little like this (only 79 and 81 are 2.0-- 80 and two other 1.9 nodes will get fresh installs once I get everything working well and migrate what's running on them):
In addition, as per an old thread, I've configured udp ports 5404 and 5405 configured as such:
However, most of the traffic I saw using tcpdump was directed at 239.192.6.89 (the address reported by pvecm status). I can't really tell anything is down from any commands, I don't see much useful information in the logs (mostly seemingly useless lines stating "Retransmit List").
We ran into some issues with the 2.6.32-11 kernel, so we're running the 2.6.32-10 (probably will report on the issue later, probably my own fault). The problem existed on the newer one when I ran it.
So, a few questions questions:
How should the host firewall be configured to allow the Proxmox cluster information.
Is there some way I can see/debug the status asides from looking for a red light in the web interface? Perhaps something to enable quicker testing of the firewall configuration?
Does the multicast address change if the entire cluster is shut down and brought back up?
Thanks,
- Alex
Initially, after rebooting the nodes, everything works and looks fine. After about 10 minutes (sometimes more), it seems to stop working until I disable the firewall (then it shows green after a few seconds to a few minutes).
I have every node configured to accept ALL traffic from all the other nodes. It looks a little like this (only 79 and 81 are 2.0-- 80 and two other 1.9 nodes will get fresh installs once I get everything working well and migrate what's running on them):
Code:
# Allow Proxmox VE Cluster Connection
-A INPUT -s xx.xxx.64.79 -j ACCEPT
-A INPUT -s xx.xxx.64.80 -j ACCEPT
-A INPUT -s xx.xxx.64.81 -j ACCEPT
In addition, as per an old thread, I've configured udp ports 5404 and 5405 configured as such:
Code:
# Allow Proxmox VE multicast traffic
-A INPUT -p udp --dst 239.192.12.206 --dport 5404 -j ACCEPT
-A INPUT -p udp --dst 239.192.12.206 --dport 5405 -j ACCEPT
However, most of the traffic I saw using tcpdump was directed at 239.192.6.89 (the address reported by pvecm status). I can't really tell anything is down from any commands, I don't see much useful information in the logs (mostly seemingly useless lines stating "Retransmit List").
We ran into some issues with the 2.6.32-11 kernel, so we're running the 2.6.32-10 (probably will report on the issue later, probably my own fault). The problem existed on the newer one when I ran it.
So, a few questions questions:
How should the host firewall be configured to allow the Proxmox cluster information.
Is there some way I can see/debug the status asides from looking for a red light in the web interface? Perhaps something to enable quicker testing of the firewall configuration?
Does the multicast address change if the entire cluster is shut down and brought back up?
Thanks,
- Alex