After upgrade from 5.1 to 5.2 corosync failed to start due to DNS lookup failure

[pero]

Hi!

I've run into a very strange upgrade process. I'm using proxmox for a very long time ( from 1.4 ).
Our DNS server was always a guest machine on one of the proxmox nodes.

Now I upgraded all nodes from 5.1 to 5.2, but the node with the DNS guest failed to join the cluster after reboot.
In about an hour I realized that this node tries to join the cluster using DNS names, but DNS guest server is not started before quorum is OK.
So I was able to start corosync by adding all node's IP/domain pairs to /etc/hosts on all nodes.

In the last 9 years I restarted the node with DNS server guest several times, and I've never ever had issues because of this. Something has changed since 5.2 in the cluster communication?

I think that joining the cluster should not depend on DNS resolving.
But if it needs DNS resolving, than all hosts file should contain all node's IPs.

Can somebody please clarify this issue in my mind?

Thank you

 

[Klaus Steinberger]

It is always a good idea to pack the necessary host adresses for booting into /etc/hosts

booting a node with its dns on its own guest is always a game of henn and egg (who is first ?)

 

[pero]

So you mean, that Proxmox needs DNS resolving after a node is rebooting?
When I join a node with pvecm, I use IP address, not domain names. I don't understand why it needs DNS resolving after reboot?
Yes, my DNS server is a guest on a node. But what if my DNS would be the DNS of my ISP, if there was an internet failure, i could not reboot my proxmox nodes?
I can accept this, that it needs resolving, but then proxmox should add all nodes to all hosts files automatically, right?
It's a bit weird that I have to edit hosts files manually. If I have a big cluster, and I add a new node to it, than i have to edit the hosts files on all nodes manually?
I'm a bit lost, knowing that I used proxmox for years without this kind of errors...
It's no problem, I can edit this files, i just want to know why i have to, and why it is not documented, that I have to