Cluster IP migration

liane

Renowned Member
Nov 25, 2008
40
1
73
Hi,


I need to change IPs of both nodes of a 2 nodes (up to date) unicast cluster with no downtime of any VM. (old IPs: 88.x.x.x, new ones: 195.x.x.x)


I planned to:
- move all VMs from one node (B) to the other (A)
- change B IP
- migrate back all VMs from A to B
- change A IP
- dispatch VMs like they were at the beginning


So far so good, I tried the following:
- move all VMs from node B to node A
- change IP of node B in
/etc/network/interfaces
/etc/hosts
also change /etc/hosts in node A
- /etc/init.d/networking restart (node B)


this led to node B glowing red in the cluster, and no VM migration possible.


- reboot node B


not better


- restore previous network config in the previously changed files
- /etc/init.d/networking restart (node B)


cluster still down


- reboot node B


cluster finally restored, back to start.


Since the new IPs are in different ranges than the old ones, I’m not sure if I have any chance to make a temporary cluster with node A in the old range and node B in the new one.


Removing node B from the cluster, moving IPs of both nodes and setting again node B in the cluster doesn’t feel that safe, according to:
http://pve.proxmox.com/wiki/Proxmox_VE_2.0_Cluster#Remove_a_cluster_node


What is the safest way to operate this migration?
 
Code:
1) You cannot change IP of a host without rebooting.2) You cannot migrate VM's to a host which is not known by the source host.
QeD: You cannot change cluster wide IP's on a two-node cluster. It would be possible when the cluster consists of at least 3 nodes.
 
Last edited:
What is the safest way to operate this migration?

the safest, and probably the fastest: backup all vm/ct, check restore of all, reinstall cluster with new nodes ip, restore storage, restore vm/ct.
otherwise, ...there are so many steps involved... and is not easy anyway.

Marco
 
thanx for your replies...

QeD: You cannot change cluster wide IP's on a two-node cluster. It would be possible when the cluster consists of at least 3 nodes.

I don't get it why it would be possible with 3 nodes when it's not with 2? Could you please explain or point me to a reference?


the safest, and probably the fastest: backup all vm/ct, check restore of all, reinstall cluster with new nodes ip, restore storage, restore vm/ct.

Of course, reinstall from scratch would work... except that it defeats the entire point of migrating without VM downtime...
 
I don't get it why it would be possible with 3 nodes when it's not with 2? Could you please explain or point me to a reference?
With 3 nodes you will be able to keep two nodes online at the same time to migrate between at any time while the third node is rebooting.
 
sorry to be so dense, but one node online should be enough in this case, I wonder what went wrong with my tries :/
 
since I'm going to try again on Sunday, just wondered if someone could have the slightest idea on how to proceed safely, before I mess up everything