Kubeflow

This guide presents an overview of the Charmed Kubeflow (CKF) charms that provide Prometheus monitoring metrics.

All metrics can be accessed using the Prometheus or Grafana User Interface (UI). See Integrate with COS for more information.

Argo controller

See the argo-controller upstream documentation for more information on provided metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="argo-controller"}

Dex auth

The dex-auth charm provides:

• A custom metric counting HTTP requests. See its source code for more details.
• Go runtime and process metrics for monitoring the controller.
• gRPC server metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="dex-auth"}

Envoy

The envoy charm provides the following metrics:

• Listener.
• HTTP connection manager.
• Cluster manager.
• Server.
• Access logs.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="envoy"}

Istio pilot

See the istio-pilot upstream documentation for more information on provided metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="istio-pilot"}

Istio gateway

See the istio-gateway upstream documentation for more information on provided metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="istio-gateway"}

Jupyter controller

The jupyter-controller provides the following metrics:

• Custom notebook-related metrics. See its source code for more details.
• Go runtime metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="jupyter-controller"}

Katib controller

The katib controller provides the following metrics:

• Custom experiment-related metrics. See its source code for more details.
• Go runtime metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="katib-controller"}

Kfp api

The kfp-api provides the following metrics:

• Custom metrics related to its several components. See its source code for more details:
• Resource manager.
• Experiment server.
• Job server.
• Pipeline server.
• Pipeline upload.
• Run server.
• Go runtime and process metrics for monitoring the controller.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="kfp-api"}

Knative eventing

The knative-eventing metrics come from the knative-operator charm that deploys otel-collector. See its upstream documentation for more details.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="knative-operator", namespace_name="knative-eventing"}

Knative serving

The knative-serving metrics come from the knative-operator charm that deploys otel-collector. See its upstream documentation for more details.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="knative-operator", namespace_name="knative-serving"}

Knative operator

See the knative-operator upstream documentation for more information on provided metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="knative-operator"}

Metacontroller operator

The metacontroller-operator provides the following metrics:

• Custom metrics. See its source code for more details.
• Go runtime and process metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="metacontroller-operator"}

Minio

See the minio upstream documentation for more information on provided metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="minio"}

Seldon controller manager

See the seldon-controller-manager upstream documentation for more information on provided metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="seldon-controller-manager"}

Training operator

The training-operator provides the following metrics:

• Custom job-related metrics. See its source code for more details.
• Go runtime and process metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="training-operator"}

Pvcviewer operator

The pvcviewer-operator provides the following metrics:

• Go runtime and process metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="pvcviewer-operator"}

Kserve controller

The kserve-controller provides the following metrics:

• Go runtime and process metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="kserve-controller"}

Kubeflow profiles

Kubeflow profiles manage two Pebble services:

• profile-controller.
• kfam.

Profile controller

The profile-controller provides the following metrics:

• Custom job-related metrics. See its source code for more details.
• Go runtime and process metrics for monitoring the controller.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="kubeflow-profiles"}

Kfam

The kfam provides the following metrics:

• Custom job-related metrics. See its source code for more details.
• Go runtime and process metrics for monitoring the controller.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="kubeflow-profiles"}

Tensorboard controller

The tensorboard-controller provides the following metrics:

• Go runtime and process metrics for monitoring the controller.
• Controller runtime metrics.

You can check its metrics through the Prometheus or Grafana UI using the following query:

{juju_charm="tensorboard-controller"}

---