update prometheus config for 2.1

[martell]

What this PR does / why we need it:
fixes #2142, as a follow up to #2183

[trstringer]

@martell (or @dtzar if you've tested this pr out), what api-model did you use to test this out successfully? Can you share?

[trstringer]

I'm unsure how this could have been tested successfully without changing this line of code in the fork.

[trstringer]

Update: we are looking into this, please do not merge this yet.

cc // @ritazh

[dtzar]

There is no way I see this fixing the entire solution since in the example Readme we do not turn off RBAC (which is now enabled by default in acs-engine). I approved just because this gets us closer to working. The best solution IMO would be to have a v1 version of the extension which works with non-RBAC clusters where the helm version will match roughly the supported K8s version, using the 4.x version of the prometheus chart. v2 version of the extension should have a switch for RBAC, but leave it on by default and target 1.8/1.9+ clusters with the 5.X version of prometheus chart.

Would be good to have a complete working solution, so taking back my approve.

[dtzar]

I just tested with a vanilla acs-engine configuration (no prometheus/grafana extension) which provisions K8s 1.7.9 configuration and it uses RBAC. Without specifying rbac.create=true the installation fails, but with this set, the installation succeeds. @martell can you please update this to be true in your PR for the values.yaml file?

[trstringer]

@dtzar you're talking about the manual helm installation of these charts succeeding? This is going to be a different case. This embedded prometheus_values.yaml config file is meant for the v4 version of the prometheus chart. It needs to be updated prior to merging.

Without RBAC set for the cluster, what is the failed error message when you try to install v5 of the Prometheus helm chart?