inventory two potential solutions

Hey guys!
We are developing a plugin for openstack horizon and a product which will be able to manage openstack infrastructure
Our infrastructure is quite complex.
For instance, we may have up to 10 openstack controllers, up to 5 000 compute nodes, and few different servers,

System administrators will be able to add/remove to our product new compute nodes, while installation of our producct on compute node, we need to some collectd plugins, etc
I’ve discussed the problem with guys in our team who knows ansible better than me.
I disagree with their solution: their proposal is to keep compute nodes, in ansible facts, so, when user adds a host we update a fact file in our product directory

At the moment we keep our compute nodes in filesystem, in some file

I’d like to go different way (at least two options):

  1. use dynamic inventory like for amazon in examples (list of compute nodes always available from file in filesystem)
  2. add dns to our infrastructure and group all hosts, and use these groups in ansible playbooks

What is the besst solution according to your experience ?

For me it’s clear that storing list of our servers in ansible facts is a wrong way.

Actually, I thought a bit more about the problem and I guess guys in our team are right about storing inventory in ansible facts
If we continue development with this huge dependency, list of our servers in specific files in os, we won’t be able to change later HOST of ansible playbooks executions, it would be hard because we have such dependency, and it would be hard to migrate to ansible tower.

понеділок, 12 лютого 2018 р. 23:01:08 UTC+1 користувач Sergey Glazyrin написав: