Quickstart

Connect your first cluster, create an alert rule, and receive a Slack notification — in about 5 minutes.

Prerequisites

• A running Kubernetes cluster (any distribution)
• Helm 3 installed and configured to access your cluster
• kubectl access with cluster-admin or equivalent
• A SpecLayer account — get started free

Step 1 — Get your API key

In the SpecLayer app, go to Settings → API Keys and create a key with the klusteralert:write scope.

Step 2 — Add the Helm repo

helm repo add speclayer https://charts.speclayer.net
helm repo update

Step 3 — Install the agent

helm install klusteralert speclayer/klusteralert \
  --namespace klusteralert \
  --create-namespace \
  --set apiKey=$SPECLAYER_API_KEY \
  --set clusterName=production

Step 4 — Verify the agent is connected

Check the agent pods are running:

kubectl get pods -n klusteralert

# Expected output:
NAME                                    READY   STATUS    RESTARTS   AGE
klusteralert-agent-6d9f8b7c4d-xkp2j    1/1     Running   0          45s

In the SpecLayer app, go to KlusterAlert → Clusters. Your cluster should appear as Connected within 30 seconds.

Step 5 — Create your first alert rule

Create an alert rule file and apply it to your cluster via ConfigMap, or create it in the UI:

apiVersion: klusteralert.speclayer.net/v1
kind: AlertRule
metadata:
  name: crash-loop-detect
  namespace: klusteralert
spec:
  description: Alert when a pod enters CrashLoopBackOff
  severity: critical
  condition:
    type: pod.status
    equals: CrashLoopBackOff
  notify:
    - channel: slack
      target: '#incidents'
  runbook: https://runbooks.acme.com/crashloop

kubectl apply -f alert-rules.yaml

Step 6 — Connect Slack

In the SpecLayer app, go to KlusterAlert → Notification Channels → Add channel. Select Slack and follow the OAuth flow to authorise the SpecLayer app in your workspace. Once connected, select the channels you want to route alerts to.

Step 7 — Test the alert

Trigger a test notification from the Notification Channels settings page, or deploy a broken pod to verify end-to-end:

kubectl run crasher --image=busybox -- /bin/sh -c "exit 1"

# Within ~30 seconds KlusterAlert fires the alert.
# Check your Slack channel for the notification.
kubectl delete pod crasher

What's next