Identify which resources need to be monitored and convert them into payloads that can be sent to Prometheus (or maybe Influxdb?) (e.g. CouchDB, nginx, iptables, Docker, host OS, Kubernetes, etc.)
Configure clients to send data and server to store/display/report on it.
Define initial operational thresholds for normal, warning and critical levels.
Define alerts in Prometheus.
Annotate events so that context is available when viewing metrics/alerts. Aid in troubleshooting.