Closed darahayes closed 6 years ago
cc @wtrocki
@darahayes Verified locally. It works like dream for general dashboard.
@paolobueno suggested splitting the grafana and prometheus provisioning into separate roles. While I initially agreed with this approach, I did some investigation and I've concluded there's no benefit in doing this for the time being.
The ansibleplaybookbundle org has a number of example APBs that provision multiple services e.g. (Application + Database) and they seem to keep them in the same roles.
@paolobueno @wtrocki what do you think?
@darahayes Do you need me to verify that again?
Yes please @wtrocki . Now that the name has been changed I think it would be wise to reverify.
@paolobueno Can you verify that as well? I'm happy to merge (already approved this change)
@wtrocki tried verfiying but blocked on my local origin cluster: https://gist.github.com/paolobueno/c3430125b2ab43f22884597f6314b29d
Feel free to go ahead and merge!
@paolobueno Looks like missing selinux options for openshift containers.
Any non trivial PR's should have 2 approvals. Added more people to share ideas.
@pb82 thanks for help with verification. Let's merge this.
The result of this PR is that grafana and prometheus are split into their own pods.
In addition, a new service called
prometheus-internal
was added which directly exposes the prometheus container. (no proxy). This additional service is needed so grafana can access prometheus as a datasource. This approach was outlined in the metrics proposal. I'm looking for suggestions on the name though.The datasource config file seems to be broken at the moment, so you will not see the prometheus data source show up automatically when the apb is run. You can manually add the prometheus service afterwards using the hostname
http://prometheus-internal
I'm aware that there is still plenty of room for improvement in this APB which I think should be tackled in a separate PR.