Hi, I’m building a HA cluster of OpenNebula frontends and I’d like to keep the hostnames on the frontend nodes. as ‘one-2’, ‘one-3’, ‘one-1’
Is there a feature to change the way hypervisor nodes refer to the frontend node? so that I can keep the hostname of the frontend nodes in the HA cluster unchanged while there is a V-hostname-thingy and a V-IP address (192.168.255.237 one one.petacnc.com) that are used on hypervisor nodes. So that in the event of shutting down a failed frontend node and running OpenNebula on another frontend node, hypervisors don’t need any update.
it seems it was all set up automatically in the first place and I just needed to put ‘192.168.255.240 one-2 one-2.petacnc.com’ in /etc/hosts on every node.
Could anybody tell me when and how the method is declared with which hypervisor nodes initiates data communication with the frontend node? I could just set a DNS record of “one-2” for example. I thought it would be the second best plan.
AFAIK the monitoring in the hypervisors is managed by the active front-end. So in your case the second front-end should restart the monitoring pointing the opennebula’s collectd service on the hypervisors to it’s own IP daddres. If I am not wrong there was a bug related to this case and the temp solution is to kill the collectd process on the HVs and reload/restart the monitoring. You should digh here in the forum for the description of the problem and the proposed solution.
Though I just found out that OpenNebula developers are working on native high availability feature in oned itself which is easy to set up and smooth in recovery phase. ( I read the April OpenNebula newsletter just now on May 13 haha )
I wonder how long it would take finishing the tests and releasing it publicly. does anyone have an idea?
AFAIK first beta should be released any moment - the developers are pushing the feature branches to the master branch and resolving the merging conflicts though.