iblancasa / kubecon24-kepler-knative

Repo with assets to reproduce the talk

kubecon24-kepler-knative

Repo with assets to reproduce the talk

Configuring OpenShift local

• crc setup
• crc start --cpus 7

Install kn cli

• wget https://mirror.openshift.com/pub/openshift-v4/clients/serverless/latest/kn-linux-amd64.tar.gz -O my-kn.tar.gz
• tar -xf my-kn.tar.gz
• sudo mv kn-linux-amd64 /usr/local/bin/kn

Install serverless

1. Install the operator by following the official doc

   • oc create -f yamls/serverless-operator.yaml

2. Create Serving

   • oc project openshift-serverless
   • oc create -f yamls/serving.yaml
   • Check: oc get knativeserving.operator.knative.dev/knative-serving -n knative-serving --template='{{range .status.conditions}}{{printf "%s=%s\n" .type .status}}{{end}}'

3. Service

   • oc create project mock-ns
   • kn service create mock --namespace mock-ns --image ghcr.io/jgomezselles/kubecon24/server-mock:0.0.1 --concurrency-limit 4 --force
   • Need to attack mock-hermes.apps-crc.testing:80

4. Create user-workload-monitoring (Prometheus)

   • oc create -f yamls/cluster-monitoring-config.yaml

5. Install power monitoring

   • Follow instructions

6. Deploy grafana

   • Simply run deploy-grafana.sh

7. Install hermes

   • helm install hermes -n mock-ns charts/hermes-job/

8. Install OTel

   • oc create -f otel-collector/
   • Check by oc get csv -n openshift-opentelemetry-operator

9. Create service monitor

   • oc create -f otel-service-monitor.yaml

Queries:

• avg(rate(hermes_requests_sent_total{id="post1"}[1m]))
• kepler:kepler:container_joules_total:consumed:24h:by_ns{container_namespace=~"$namespace"} * $watt_per_second_to_kWh

Building mock image

• docker build -f server-mock/docker/Dockerfile . -t ghcr.io/jgomezselles/kubecon24/server-mock:0.0.1 --progress plain --no-cache

Deleting installation

• kn service delete mock
• oc delete -f yamls/serving.yaml
• oc delete -f yamls/serverless-operator.yaml

Doubts

• Internal endpoint?

Next

• Update helm and installations with https://jgomezselles.github.io/hermes-charts/
• Deploy mock as server
• Add OTel collector to helm charts?
• Update namespaces in svc monitor and collector
• Find the best way to deploy the mock with a LoadBalancer in http/2
• Install both at the same time and run
• Create dashboard
• See if we can save prometheus metrics to just reuse them

Fix

• Fix service monitor in the otel collector
• Add endpoint to hermes to export OTLP metrics