Add Prometheus rules to swan-cern chart

This is done so that they are deployed everytime whenenever
the swan-cern chart is deployed.
They define rules that are evaluated every 24h to create metrics
about SWAN unique users. The ones that contain "list" in the name
generate a list of unique users and the metrics that contain "peak"
calculate the total number of unique users in that day (both normal
and GPU users).

swan-cern/templates/prometheus/rules.yaml

@@ -0,0 +1,21 @@
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  labels:
+    app: prometheus
+    release: cern-magnum
+  name: swan.rules
+  namespace: {{ .Release.Namespace }}
+spec:
+  groups:
+  - interval: 24h
+    name: swan.rules.users
+    rules:
+    - expr: count(last_over_time(kube_pod_status_phase{namespace="swan",phase="Running",pod=~"jupyter-[a-z]+"}[24h])>0) by (pod)
+      record: swan:users:unique:list:daily
+    - expr: max_over_time(count(count(kube_pod_status_phase{namespace="swan",phase="Running",pod=~"jupyter-[a-z]+"}>0) by (pod))[24h:5m])
+      record: swan:users:peak:daily
+    - expr: count(last_over_time(kube_pod_container_resource_requests{resource=~"nvidia_com_(gpu|mig.+)"}[24h])) by (pod)
+      record: swan:users:gpu:unique:list:daily
+    - expr: max_over_time(count(count(kube_pod_container_resource_requests{resource=~"nvidia_com_(gpu|mig.+)"}) by(pod))[24h:5m])
+      record: swan:users:gpu:peak:daily