Auto Import

Auto import reads port-forward configurations from Kubernetes service annotations, so you don't need to create them by hand. The configs stay in sync with the service definitions.

Annotation formatCopy Anchor Link

apiVersion: v1
kind: Service
metadata:
  name: monitoring-stack
  namespace: observability
  annotations:
    kftray.app/enabled: "true"
    kftray.app/configs: "grafana-3000-3000,prometheus-9090-9090,alertmanager-9093-9093"
spec:
  selector:
    app: monitoring
  ports:
    - name: grafana
      port: 3000
    - name: prometheus
      port: 9090
    - name: alertmanager
      port: 9093

The kftray.app/configs annotation defines port forwards in the format alias-local_port-remote_port. Separate multiple forwards with commas. The alias and a Service port name can contain hyphens, as in my-api-8080-http-alt.

This example creates three configurations. Each one forwards localhost:<local_port> to the service's container port.

Remote port namesCopy Anchor Link

remote_port can also be a Service port name. kftray then forwards to that port's targetPort on the pod, not to the Service's own port.

annotations:
  kftray.app/enabled: "true"
  kftray.app/configs: "api-8080-http,admin-8081-admin"
spec:
  ports:
    - name: http
      port: 80
      targetPort: 8080
    - name: admin
      port: 8080
      targetPort: 9090

Here api forwards localhost:8080 to container port 8080, and admin forwards localhost:8081 to container port 9090. A number in the annotation is used as given, so the Service's exposed port is never translated a second time.

A named targetPort is resolved from the container ports of the pods the Service selects. Those pods must declare the same number for that name. If they disagree or kftray can't list them, it skips the port and logs a warning.

How discovery worksCopy Anchor Link

kftray scans services in every namespace you can access. It skips a namespace it can't list and logs the failure. For each service with kftray.app/enabled: "true", kftray:

Discovery runs when you start it, from the desktop menu or from kftui. kftray doesn't rescan on a timer.

Integration with HelmCopy Anchor Link

# templates/service.yaml
apiVersion: v1
kind: Service
metadata:
  name: {{ include "myapp.fullname" . }}
  annotations:
    kftray.app/enabled: "true"
    kftray.app/configs: "{{ .Values.kftray.alias }}-{{ .Values.kftray.localPort }}-http"
spec:
  ports:
    - name: http
      port: {{ .Values.service.port }}
      targetPort: 8080

# values.yaml
service:
  port: 80

kftray:
  alias: "myapp"
  localPort: 8080

Deploy with Helm, then run auto import to pick up the service:

helm install myapp ./myapp
# auto import creates: myapp-8080-http

Using auto importCopy Anchor Link

Imported configurations merge with existing ones, so running auto import several times doesn't create duplicates.