Set grafana collectd need to make visibility of all of our clusters

asked 2018-09-03 04:29:50 -0600

Steve987654321 gravatar image


Let me define first what is my goal:
I want to have pretty Grafana dashboards about our openstack clusters. We have 5 datacenters with around 3-4000 physical machine and 15k vm-s. My task is to create some pretty Grafana dashboards fo mysql things, rabbit mq things and of course about the openstack performance, instance counts … We want to use it for capacity planning os something like that, for example in the last 1 year how many xxlarge vm-s were requested in each tenant.

The last week I read a lot about this because I completely new in this area and now I completely got lost regarding the objects, gnocchi, ceilometer, telemetry, aodh, collectd, Prometheus, influxdb … so in my head at the moment I have a completely big mess.

Our current Grafana infrastructure is actually built, I should integrate our openstack stuffs into it.

  1. Collectd is running everywhere
  2. Graphite is collecting the collect things
  3. Grafana is reading the graphite

I can modify, add more source to the graphite server, but please help me, what should I do, which is the best was to make it work?

I’ve tried dashboards imported from Grafana but I can’t make it work, it always show idiot information.

So how is should collect the data? Gnocchi or collectd or something else?

If I collect with gnocchi do I have to add new source for graphite where we have the data?

Do I need aodh or ceilometer (telemetry)?

Our DC-s looks like the following:

  • 2 Controller nodes running neutron, nova, keystone
  • Separated load balancers
  • separated rabbit mq
  • separated db servers
edit retag flag offensive close merge delete