Get Your First Slack Alert¶

Break a pod, watch it enter CrashLoopBackOff, and confirm alertkube sends Slack.

Prerequisites

You've already installed alertkube and confirmed it's healthy in Install alertkube with Helm in 5 minutes. You can post to a Slack channel via webhook or bot token.

Configure Slack¶

If slack.webhookUrl was set during install, skip to the pod test.

helm upgrade --install alertkube oci://ghcr.io/aryasoni98/charts/alertkube --version 1.2.0 \
  --set cluster=my-cluster \
  --set slack.webhookUrl=https://hooks.slack.com/services/Change-Me

Use bot-token mode for real per-severity channels. The bot needs chat:write and must be invited to each channel.

helm upgrade --install alertkube oci://ghcr.io/aryasoni98/charts/alertkube --version 1.2.0 \
  --set cluster=my-cluster \
  --set slack.botToken=xoxb-your-bot-token \
  --set slack.channels.critical=alerts-critical \
  --set slack.channels.warning=alerts-warning \
  --set slack.channels.info=alerts-info

When a bot token is set, it takes precedence over the webhook URL.

Break a Pod¶

kubectl run boom --image=busybox --restart=Always -- /bin/false

Watch it:

kubectl get pod boom -w

Wait for CrashLoopBackOff, then stop watching.

Read the Alert¶

The message includes severity, resource identity, reason, fingerprint, and optional runbook link. Repeats are muted by fingerprint for behavior.muteSeconds.

Add a runbook link:

kubectl annotate pod boom runbook-url=https://wiki.example.com/runbooks/crashloop

Clean Up¶

Delete the broken pod:

kubectl delete pod boom

After the fingerprint is quiet for behavior.resolveTTLSeconds, alertkube sends a resolved alert. For tests, the TTL can be lowered but must stay above 300 seconds.