Is kwatch working?

kwatch runs as one pod. There is no high availability in v1: if that pod is down, kwatch is not watching until Kubernetes restarts it. So it is worth knowing how to check that it is alive, and that its alerts can reach you.

Check it yourself

1. Run the installer and choose View status. It shows the pod, the version and the license state. The installer can also show logs.

2. Choose Send test alert. It sends a clearly marked test through your real notification channel, so a pass proves the same path real alerts use.

3. Prefer the command line? The same test runs inside the pod:

   kubectl -n <namespace> exec deploy/kwatch -- kwatch test-alert

   Add --provider <name> to test one provider. A test alert is not stored as an incident.

If the test does not arrive, see the silent failures below.

Cluster check-ins Pro

With an active Pro license, kwatch checks in with api.kwatch.dev every 30 minutes. app.kwatch.dev shows each cluster's last seen time and its state:

State	Meaning
Checking in	kwatch checked in recently.
Silent	Nothing heard for 60 minutes. kwatch stopped talking; your apps may be fine, but kwatch cannot alert you.
Stopped	kwatch shut down cleanly, for example scaled down on purpose. Not a problem.
Check-ins off	You turned them off.

Nothing is sent to anyone: no email, no message. You look at the page. A crash, an out-of-memory kill, a lost node or blocked outbound access shows as Silent after 60 minutes.

Community sends no check-ins. To turn them off, open the installer's License / Pro menu and choose Cluster check-ins.

Your own heartbeat URL Pro

If you want a message when kwatch goes quiet, point kwatch at a monitor you own. See Heartbeat monitor.

Common silent failures

• A provider credential is wrong or expired. The test alert reports the rejected credential. Fix it in Edit notification providers.
• Community supports one provider. Extra providers are ignored. See Community vs Pro.
• Namespace scope. kwatch only watches the namespaces it is configured for. Check the namespace settings if one workload never alerts.
• A NetworkPolicy blocks traffic. kwatch needs outbound access to your provider (and to api.kwatch.dev for Pro), and to port 10250 on the nodes for node stats.
• The startup baseline. After a start kwatch learns what is already failing and sends one summary instead of announcing every old problem. Only new or worsening problems alert.

Still stuck? See Getting help.