First release - prometheus-exporter operator

[slopezz]

Solves internal issue https://github.com/3scale/platform/issues/42

Main features

• It has been added first version of prometheus-exporter-operator using operator-sdk v0.13.0 (last available version)
• At the moment it is an ansible based operator, but the idea is to migrate it to go in the future
• This operator can be used on Kubernetes or OpenShift
• Current supported prometheus-exporter type: memcached, redis, mysql, postgresql, sphinx, elasticsearch, cloudwatch
• Operator manages lifecycle of the following objects for every CR: Deployment, Service and optionally ServiceMonitor
• Some exporters require previously manually created objects like Secrets for credentials and Configmaps for configuration files

Operator

• For operator image management, you will need operator-sdk binary, you can get it here
• To build/push an operator image just run (takes image repository and version from Makefile):

  make operator-image-update

• To deploy operator objects (Namespace, CRD, role, role_binding. service_account, operator deployment):

  make operator-create

• To clean resources:

  make operator-delete

• Deletion of CRD and Namespace is not integrated on the Makefile delete targets for caution:
  • Deletion of Namespace may delete undesired objects on the Namespace that later on may not be reproducible
  • Deletion of CRD (which is an object at cluster level) may caused other possible deployed CRs on different cluster Namespace to fail

Exporters examples

• On examples directory there is included a secondary Makefile used to deploy examples of each supported prometheus-exporters (including extra needed objects like Secrets/Configmaps, example DBs, example CRs..) and also to manage created objects deletion.
• All create/delete targets are documented on examples/README, or can be checked with make help
• All examples of exporters monitoring databases include the deployment of an example database (except sphinx or elasticsearch, where you will need to set a previously available/reachable dbHost/dbPort on its pertinent CR)

GrafanaDashboard/PrometheusRule

• It has been added examples of grafana dashboards for memcached, redis, mysql, postgresql, sphinx
• In addition, it has been added examples of prometheus rules for memcached, redis, mysql, postgresql, sphinx (alert thresholds may need to be customized for each case, depending on databases size)
• In order to test grafana-dashboards and prometheus-rules, it has been used https://github.com/integr8ly/application-monitoring-operator which includes prometheus, alert-manager and grafana operators deployment on Namespace application-monitoring
• To test them, you just need to be authenticated within a Openshift v.3.11.0+ cluster, and:

  
  git clone https://github.com/integr8ly/application-monitoring-operator.git

make cluster/install

- On Namespace `application-monitoring` you will see all created resources, and among them you will see public `Route` created objects for Prometheus, Grafana and AlertManager.
- For managing creation/deletion of dashboards/rules, you can use Makefile targets documented on README:

make grafana-dashboards-create make grafana-dashboards-delete make prometheus-rules-create make prometheus-rules-delete

[roivaz]

I have a comment on the directory structure. Why have a deploy dir and then an examples dir? I think is more easy to a potential user if all were under the deploy directory, which is usually what you find in other operator projects.

Same can be said about the grafana-dashboards one, I would include everything under deploy