Hi
I have 15 nodes running proxmox VE 6 and CEPH. Recently we've been experiencing unexpected reboot in one or two nodes and logs showed that their corosync couldn't communicate with each other, therefore the node decided to fence itself and do a reboot automatically.
Even though we had two separated networks, 10Gb for network and 100Gb for CEPH, we decided to add a 1Gb network for corosync to eliminate the possibility of having high latency in the corosync.
Now I'm planning on separating the corosync network without downtime or a minimum interruption.
According to this article :
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_cluster_network
We will setup the network interfaces and make sure that they can see each other, then edit the corosync.conf file.
The file says there might an interruption as each node needs to refresh or restart corosync to get sync in a new network, but I'm worried if any of the nodes get rebooted like before for any unknown reason.
So in order to prevent this, can I stop "ha" service in all nodes before doing this? so even if the nodes couldn't join make quorum after the change, none of them get rebooted?
And another question.
My corosync.conf file has a is a bit different than the one shown in the article.
in the totem part, I have this:
totem {
cluster_name: proxmoxcluster1
config_version: 24
interface {
bindnetaddr: xx.xx.xx.xx <<<< ( This is the IP of one of my first nodes)
ringnumber: 0
}
I can see that there is no "bindnetaddr" in the example in the article. So can I safely remove this line.
What does it do anyway?
I would appreciate any help on this.
I have 15 nodes running proxmox VE 6 and CEPH. Recently we've been experiencing unexpected reboot in one or two nodes and logs showed that their corosync couldn't communicate with each other, therefore the node decided to fence itself and do a reboot automatically.
Even though we had two separated networks, 10Gb for network and 100Gb for CEPH, we decided to add a 1Gb network for corosync to eliminate the possibility of having high latency in the corosync.
Now I'm planning on separating the corosync network without downtime or a minimum interruption.
According to this article :
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_cluster_network
We will setup the network interfaces and make sure that they can see each other, then edit the corosync.conf file.
The file says there might an interruption as each node needs to refresh or restart corosync to get sync in a new network, but I'm worried if any of the nodes get rebooted like before for any unknown reason.
So in order to prevent this, can I stop "ha" service in all nodes before doing this? so even if the nodes couldn't join make quorum after the change, none of them get rebooted?
And another question.
My corosync.conf file has a is a bit different than the one shown in the article.
in the totem part, I have this:
totem {
cluster_name: proxmoxcluster1
config_version: 24
interface {
bindnetaddr: xx.xx.xx.xx <<<< ( This is the IP of one of my first nodes)
ringnumber: 0
}
I can see that there is no "bindnetaddr" in the example in the article. So can I safely remove this line.
What does it do anyway?
I would appreciate any help on this.