Monitoring with Prometheus doesn't start

[Spyros Trigazis]

With kubernetes running in containers, the deployment time is a bit longer.

We create the monitoring stack from the master
- we do requests to the apiserver to create the deployment for
  prometheus and grafana while there are no nodes available
- we wait for grafana to be running to inject the prometheus datasource

Because the node comes up a little later now, kubernetes seem to not
schedule the grafana and prometheus pods.

We may need to check when a node is ready and then create the monitoring stack
OR
We may need to create from a node, where we are sure that a node is available.

[Spyros Trigazis (strigazi)]

I tested in our production environment and it works fine. I'll try a comple more times in devstack and see how it goes.

Marking as inclomplete until I can verify it.

[Launchpad Janitor (janitor)]

[Expired for Magnum because there has been no activity for 60 days.]