Metrics such as operand object counts to make sure things don’t get out of control, as well as reconcile loop performance metrics and error rates. Error rates will be very useful as some changes are supported by the kubernetes API but are not supported by the strimzi operator. We should add Prometheus metrics to our operator to allow for better monitoring.

Examples of such metrics could be:

• Number of clusters it is operating (e.g. 5 Kafka cluster, 3 Connect cluster etc.). Possibly this can be extended to something like "4 health Kafka cluster, one unhealthy" etc.
• Number of reconciliations (to see that the operator works fine)
• Number of times an operation timeouted (e.g. waiting for pod to get ready) - might indicate an error or need to increase a timeout
• Maybe some lengths of the reconciliations or rolling updates