OMD Labs Clustering - Best Practice

Hi folks,

is there a best practice approach to make OMD Labs high available?

  • Is it worth to use pacemaker?
  • How far could mod_gearman do the job?

BR
Oliver

Best practice is to use usually already available HA methods for virtual machines and put OMD into a VM. Its way easier to make the hole machine HA than every single service.

If it has to be a physical cluster, there is a german howto here: http://blog.simon-meggle.de/tutorials/nagiosomd-cluster-mit-pacemakerdrbd-teil-1/

Mod-Gearman worker is a different thing and does not replace a OMD cluster. But on the worker side, simply create two or more machines with the same Mod-Gearman configuration and you have your
cluster setup ready. The only pitfal here are plugins which write local tmp files to calculate the state in between runs. You either pin those plugins to a fixed worker or create some sort of shared filesystem.

@sni thx 4 the link, Simons page seem down - I’ll check later

the mod-gearman approach is very nice, that would cluster the “monitoring site”.
Just need a solution for core which holds manages the queue and stores the data

We make the central machine high available with virtual machine technology usually. Thats much easier than trying to make every component high available. It’s not only the core, but graphing and databases as well.
There are way to many things involved and i like to keep it simple.