charts/wire-server-metrics: Use kube-prometheus-stack

The old version of prometheus operator chart is no longer compatible with latest
K8s versions. The chart has also since moved to kube-prometheus-stack on the
prometheus-community helm chart reposistory from the helm-stable repository.

The kube-prometheus-stack helm chart also allows creating dashboard using labels
on configmaps, so now we don't have to refer to non-pinned versions of the
dashboard definitions. Instead each dashboard now becomes a configmap with a
specific label and grafana discovers it.

This commit also updates the dashboards to use new ways of addressing the pods
which came with the move from old version of prometheus-operator chart to the
kube-prometheus-stack chart. Along with these changes, the dashboards now also
use the qualified endpoints and federation endpoints to display information
about prekey claims and message sending stats.

Author: akshaymankar

Makefile

@@ -8,7 +8,7 @@ DOCKER_TAG            ?= $(USER)
 # default helm chart version must be 0.0.42 for local development (because 42 is the answer to the universe and everything)
 HELM_SEMVER           ?= 0.0.42
 # The list of helm charts needed for integration tests on kubernetes
-CHARTS_INTEGRATION    := wire-server databases-ephemeral fake-aws nginx-ingress-controller nginx-ingress-services
+CHARTS_INTEGRATION    := wire-server databases-ephemeral fake-aws nginx-ingress-controller nginx-ingress-services wire-server-metrics
 # The list of helm charts to publish on S3
 # FUTUREWORK: after we "inline local subcharts",
 # (e.g. move charts/brig to charts/wire-server/brig)

changelog.d/2-features/metrics

@@ -0,0 +1 @@
+Use kube-prometheus-stack instead of prometheus-operator and update grafana dashboards for compatibility and add federation endpoints to relevant queries.