Monitoring
telectl surfaces cluster health three ways: resource usage, events, and watching.
Top — resource usage
/top pods [-n namespace]
/top nodes
Shows CPU/memory for pods or nodes, sourced from the metrics API
(metrics.k8s.io).
- Requires metrics-server installed in the cluster. If it’s absent, telectl says “Metrics unavailable — this needs metrics-server installed” and suggests alternating rather than surfacing a raw API 404.
- Node metrics are cluster-scoped (always all nodes); pod metrics follow your current namespace.
Menu path: Monitor → Top.
Events — cluster activity
/events [-n namespace]
Lists the most recent events in the namespace: object, type, reason, and a human-readable message. Useful for spotting failed pulls, evictions, or readiness gates.
- Events expire after about an hour, so a quiet namespace may legitimately show “No events” — telectl calls this out so it isn’t mistaken for a fault.
- Menu path: Monitor → Events.
Watch — live updates
/watch <resource> [-n namespace]
Follows a resource (e.g. pods) and pushes updates as objects change. Because
it streams, it runs as a typed command, not inside a menu pane.
- Stop with
/cancel. - Output uses the session’s impersonated client, so a read-only user only sees what their role permits.
Menu entry point
Monitor (reply keyboard / menu) → pick Top, Events, or Watch. Top and Events render into the single pane; Watch shows the command syntax (it can’t live in the pane because it streams).
Grafana / dashboards?
telectl is a chat operator, not a dashboard. For long-lived metrics dashboards
pair it with your existing monitoring stack (Prometheus + Grafana, etc.);
the /top and /watch commands are for quick, on-the-go checks from chat.