Please help! I somehow messed up my corosync.conf

[silke]

I wanted to use a dedicated NIC for cluster network. First I changed /etc/network/interfaces and made sure that the new IPs were accessible from all nodes.
Then I changed corosync.conf (I thought) according to the documentation in https://pve.proxmox.com/wiki/Cluster_Manager#pvecm_edit_corosync_conf
I put the new IPs als ring0 and added the old as ring1
systemctl status now gave this error:
Sep 11 16:30:24 pve01 corosync[1613]: [TOTEM ] new config has different address for link 0 (addr changed from 192.168.178.201 to 10.1.1.1). Internal value was NOT c>
Sep 11 16:30:24 pve01 corosync[1613]: [TOTEM ] new config has different address for link 0 (addr changed from 192.168.178.202 to 10.1.1.2). Internal value was NOT c>
Sep 11 16:30:24 pve01 corosync[1613]: [TOTEM ] new config has different address for link 0 (addr changed from 192.168.178.203 to 10.1.1.3). Internal value was NOT c>
Sep 11 16:30:24 pve01 corosync[1613]: [CFG ] Cannot configure new interface definitions: To reconfigure an interface it must be deleted and recreated. A working i>

Because of the errors I copied corosync.conf.bak back to corosync.conf. But this did not restore the original state. On the web interface node 2 and 3 have red crosses and when I tried to edit again I got a permission denied:

root@pve01:~# cp /etc/pve/corosync.conf /etc/pve/corosync.conf.new
cp: cannot create regular file '/etc/pve/corosync.conf.new': Permission denied

So I guess the special shared file system is broken. How can I repair this? Please help!

 

[silke]

puh, I hope I somehow got back to my original state. This is what seemed to have helped (these three commands from another post in the forum):

systemctl stop pve-cluster
systemctl stop corosync
pmxcfs -l

Then I could edit corosync.conf and had to increment config_version (in my case to 4)
After a reboot the cluster seemed to work again.
I strongly recommend to add a recovery procedure to the documentation! In the "edit corosync.conf" section it only says to make a corosync.bak but since it is not enough to copy it back, there should be some advice how to do it (or did I just didn't see it?)

Surprisingly when I looked into corosync.conf again after the reboot it was again back to config_version 3 ! I guess that the setting of config_version 4 made it active for a short time, just to join the cluster but was overwritten by the two other nodes who were in the majority.
Whatever it seems to work again, but...

... how can I change the IPs as originally intended? The error message (s. first post) indicates that I first have to delete the old interface? That sounds strange. What do I have to do to change the IPs from the vmbr0-IPs to the new 10.1.1.1 to 10.1.1.3 and (perhaps in a second step) add the old IPs for redundency?