Metrics ports must be exposed so prometheus operator can match endpoints

[f84anton]

Prometheus operator creates scrape job with relabel_config which check if pod has exposed port.

For example, prometheus operator service monitor spec:

  endpoints:
  - targetPort: 8080
    scheme: http
  jobLabel: mysql-cluster
  selector:
    matchLabels:
      v1alpha1.mysql.oracle.com/cluster: mysql-cluster
  namespaceSelector:
    matchNames:
      - infra

creates scrape job with kubernetes_sd_configs:

  - role: endpoints
    namespaces:
      names:
      - infra
  relabel_configs:
  - source_labels: [__meta_kubernetes_service_label_v1alpha1_mysql_oracle_com_cluster]
    separator: ;
    regex: mysql-cluster
    replacement: $1
    action: keep
  - source_labels: [__meta_kubernetes_pod_container_port_number]
    separator: ;
    regex: "8080"
    replacement: $1
    action: keep

Because container ports cannot be updated statefulsets must be patched before upgrading mysql operator:

kubectl patch statefulsets.apps mysql-cluster --type=json -p='[{"op": "add", "path": "/spec/template/spec/containers/1/ports", "value": [{"containerPort":8080}] }]' 

[f84anton]

There is errors in e2e tests with "helm output: Error: could not find a ready tiller pod".