I’m in the process of migrating the Icinga 2 master nodes to new infrastructure and therefore need to move them to new hosts.
At the same time a satellite zone will be merged to the master zone.
Both the master and satellite nodes have connections to Icinga 2 agents and execute remote checks. No NRPE, no NSClient++ etc.
The current plan looks like this:
- Use Thomas’ disaster recovery tool to move all relevant files from the old to the new config master: https://github.com/widhalmt/icinga2-disaster-recovery
- Rename the names of the Icinga zones and endpoints in the director config and deploy
- Join the new second master to the new config master, creating a new master zone
- Rename the parent zone and endpoints on all agents
To avoid this in the future I’ll probably run the new masters behind generic CNAMES so I don’t have to touch all agents.
Any feedback, ideas, etc are appreciated before I move forward.
I am not experienced in migrating monitoring servers or doing a recovery, but I would say that this step should be done differently:
Renaming/changing inside the zones.conf file(s) and then importing those changes via kickstart into the Director.
At least that is how I have set up my monitoring systems
Honestly, I would create a staging environment with the current setup, and test-drive the migration strategy. Especially the need for an uptodate zones.conf in the endpoints, and then deploying with the Director should be tested well enough before hurting production.
I ended up successfully migrating to new master nodes by doing the steps in my OP except for step 2.
Like @log1c suggested I used the kickstart wizard and deleted the old endpoints and hosts.
The whole migration went smoothly, no major hickups.
it was a long time ago, but did you switch to using cnames during the migration as planned?