Skip to main content

Monitor a 3B Helm release

Written by Aron Day

BETA You’re responsible for monitoring and scaling your self-hosted 3B deployment. Use this guide at your own risk and test changes before applying them in production.

Check pod usage, disk space, and database health with your cluster’s monitoring stack. The examples assume 3B in 3b and release monitoring in monitoring. Substitute your names as needed. For a new stack, see Install a monitoring stack with Helm.

How do I access Grafana?

If your team manages Grafana or you lack cluster access, ask your operator for the private URL and sign-in. For the bundled Grafana, use the workstation where you’ll open the browser and its kubeconfig.

kubectl config current-context
kubectl -n monitoring get service -l app.kubernetes.io/name=grafana

Confirm the context and look for monitoring-grafana on 80/TCP. With permission to read the Secret, get the initial password for admin and start the port-forward.

kubectl -n monitoring get secret monitoring-grafana -o go-template='{{index .data "admin-password" | base64decode}}{{"\n"}}'
kubectl -n monitoring port-forward --address 127.0.0.1 service/monitoring-grafana 3000:80

At Forwarding from 127.0.0.1:3000, open http://127.0.0.1:3000 and sign in. Under Dashboards, open Kubernetes / Compute Resources / Namespace (Pods) and select your 3B namespace. Keep the terminal open until you’re done, then press Ctrl+C.

Check the release in the cluster

namespace=3b
kubectl -n "$namespace" get deployments,statefulsets,pvc
kubectl -n "$namespace" top pods

Look for matching ready and desired replicas, a bound Blobstore claim, and, for in-cluster Postgres, a postgres StatefulSet and its claim.

kubectl top pods shows current CPU and memory. If the metrics API is unavailable, check Metrics Server. CPU-based autoscaling also uses it.

Check coverage in your monitoring stack

In Grafana’s Explore, select your cluster’s Prometheus data source and paste this query in Code mode. It reads kubelet’s cAdvisor metrics and returns working-set memory in MiB per 3B pod. Compare the pod names with kubectl top pods.

sum by (pod) (
  container_memory_working_set_bytes{namespace="3b",container!="",container!="POD"}
) / 1024^2

Alert on used space against capacity for the data-blobstore-0 claim. For the Prometheus query and an in-pod check, see the Blobstore disk guide.

For in-cluster Postgres, track connections and disk usage on pgdata-postgres-0. For managed Postgres, check the provider’s connection, storage, and replication metrics. Postgres monitoring

Common questions

Why won’t the Grafana URL open?

With the forward running, http://127.0.0.1:3000/api/health should return "database": "ok". If port 3000 is busy, use 3001:80 and open http://127.0.0.1:3001. For NotFound, check the context, namespace, and Service name. For Forbidden, ask your operator for Grafana access.

Can I open Prometheus directly?

For the bundled Prometheus, open another workstation terminal.

kubectl -n monitoring port-forward --address 127.0.0.1 service/monitoring-kube-prometheus-prometheus 9090:9090

Open http://127.0.0.1:9090 for queries or http://127.0.0.1:9090/targets for scrape health. Press Ctrl+C when done. For managed Prometheus, use your provider’s query console or Grafana.

Why does Grafana show no pod data when kubectl top works?

kubectl top reads Metrics Server. Grafana reads Prometheus. Check the Grafana data source, namespace, and time range. Run min by (job) (up) and look for kubelet with value 1. If pod memory is missing, inspect its cAdvisor target on the Prometheus targets page.

Where can I see autoscaling activity?

After enabling 3B autoscaling, open the optional KEDA overview dashboard. Compare Worker slots wanted with 3B replicas managed by KEDA. For setup and checks, see Autoscale a 3B Helm release.

Did this answer your question?