after adding second ring adress cluster failed

[felipe]

Hi,

i tried to add a second ring address as described. Directly after i rbootet the first node the cluster lost communication. Maybe some problems in the second network with multicast and so on...
Pvecm status says allways that the nodes are in quorum but the pmxcfs fails. Rebooting a node does not help. (also i cannot reboot al nodes with running vms on it)
I tried to manually change /etc/corosync/corosync.conf on every node and restart it,,, which works,,, but if i start the cluster service it reloads the old config.
How can i revert to the old config to be sure to have the same conditions as bevore?
Is it ok to start "pmxcfs -l" on every node and change the /etc/pve/corosync.conf back and then kill it and restart corosnc and cluster services? Or does it make no effekt? Or do i get bad race conditions only changing the corosync file?

I checked that multicast is running fine (actually imgp snooping was not configured before on one switch but the cluster worked without problems) No i enabled it and multicast works good. It just seems that at the moment the firewall is blocking it. So i deactivated the firewall on the nodes (because i can not change the firewall without working cluster service)

Best regards
Philipp

 

[Stoiko Ivanov]

I tried to manually change /etc/corosync/corosync.conf on every node and restart it

did you increase the version in the corosync.conf?

Is it ok to start "pmxcfs -l" on every node and change the /etc/pve/corosync.conf back and then kill it and restart corosnc and cluster services?

that should work - although a kill -15 should be enough for pmxcfs (should because I haven't tried that recently myself, but if all corosync.conf (on all nodes both /etc/pve/corosync.conf and /etc/corosync/corosync.conf) are identical, nothing should change them)
make sure to increase the config-version! (and keep it consistent for all files)

hope this helps!