Skip to content

Controller Metrics

The Argo Rollouts controller is already instrumented with Prometheus metrics available at /metrics in port 8090. You can use these metrics to look at the health of the controller either via dashboards or via other Prometheus integrations.

Installing and configuring Prometheus

To take advantage of the metrics you need to have Prometheus installed in your Kubernetes cluster. If you don't have an existing installation of Prometheus you can use any of the common methods to install it in your cluster. Popular options include the Prometheus Helm chart or the Prometheus Operator.

Once Prometheus is running in your cluster you need to make sure that it scrapes the Argo Rollouts endpoint. Prometheus already contains a service discovery mechanism for Kubernetes, but you need to configure it first. Depending on your installation method you might need to take additional actions to scrape the Argo Rollouts endpoint.

For example, if you used the Helm chart of Prometheus you need to annotate your Argo Rollouts Controller with the following:

spec:
  template:
    metadata:
      annotations:
        prometheus.io/scrape: "true"
        prometheus.io/path: /metrics
        prometheus.io/port: "8090"

You can always see if the controller is reached successfully in the Prometheus "Targets" screen:

Prometheus Scraping Argo Rollouts metrics

Once the controller metrics are read by your Prometheus instance, you can use them like any other Prometheus data source.

Creating Grafana Dashboards

You can easily visualize the metrics from the controller using Grafana dashboards. Install Grafana in your cluster and connect it your Prometheus instance. Then you can create any dashboard by using the available metrics (described in detail in the next sections).

As a starting point you can find an existing dashboard at https://github.com/argoproj/argo-rollouts/blob/master/examples/dashboard.json

Example Grafana Dashboard

You can import this Dashboard in your Grafana installation as a JSON file.

Available metrics for Rollout Objects

The Argo Rollouts controller publishes the following prometheus metrics about Argo Rollout objects.

Name Description
rollout_info Information about rollout.
rollout_info_replicas_available The number of available replicas per rollout.
rollout_info_replicas_unavailable The number of unavailable replicas per rollout.
rollout_info_replicas_desired The number of desired replicas per rollout.
rollout_info_replicas_updated The number of updated replicas per rollout.
rollout_phase [DEPRECATED - use rollout_info] Information on the state of the rollout.
rollout_reconcile Rollout reconciliation performance.
rollout_reconcile_error Error occurring during the rollout.
experiment_info Information about Experiment.
experiment_phase Information on the state of the experiment.
experiment_reconcile Experiments reconciliation performance.
experiment_reconcile_error Error occurring during the experiment.
analysis_run_info Information about analysis run.
analysis_run_metric_phase Information on the duration of a specific metric in the Analysis Run.
analysis_run_metric_type Information on the type of a specific metric in the Analysis Runs.
analysis_run_phase Information on the state of the Analysis Run.
analysis_run_reconcile Analysis Run reconciliation performance.
analysis_run_reconcile_error Error occurring during the analysis run.

Available metrics for the controller itself

The controller also publishes the following Prometheus metrics to describe the controller health.

Name Description
controller_clientset_k8s_request_total Number of kubernetes requests executed during application reconciliation.
workqueue_adds_total Total number of adds handled by workqueue
workqueue_depth Current depth of workqueue
workqueue_queue_duration_seconds How long in seconds an item stays in workqueue before being requested.
workqueue_work_duration_seconds How long in seconds processing an item from workqueue takes.
workqueue_unfinished_work_seconds How many seconds of work has done that is in progress and hasn't been observed by work_duration. Large values indicate stuck threads. One can deduce the number of stuck threads by observing the rate at which this increases.
workqueue_longest_running_processor_seconds How many seconds has the longest running processor for workqueue been running
workqueue_retries_total Total number of retries handled by workqueue

In addition, the Argo-rollouts offers metrics on CPU, memory and file descriptor usage as well as the process start time and memory stats of current Go processes.