Skip to main content

Metrics

GoKubeDownscaler exposes a set of Prometheus metrics that can be used to monitor its activity and performance. These metrics provide insights into the targets, scaling operations, performance and general statistics and can be used to set up alerts or dashboards.

The metrics are available at the /metrics endpoint of the GoKubeDownscaler, which is exposed on port 8085.

Metrics are divided into three main categories:

  • Dry-Run Metrics: A set of metrics that are exposed only when the GoKubeDownscaler is run in dry-run mode. They are useful to monitor the potential. impact of configurations and the potential economic savings in terms of saved CPU and memory. They provide a great aid in the adoption phase of the GoKubeDownscaler.
  • Production Metrics: A set of metrics that are exposed when the GoKubeDownscaler is run in the standard mode. They provide insights into the actual. scaling operations, performance and real economic savings of the Downscaler.
  • Common Metrics: A set of metrics that are always exposed, regardless of the mode the GoKubeDownscaler is running in. These metrics provide general statistics and performance information.
tip

We recommend to set up alerts only for production metrics, as the dry-run metrics are not meant to be used in a production environment.

In the next section, we provide a detailed list of all the metrics exposed by the GoKubeDownscaler main component and by the GoKubeDownscaler Webhook

List of GoKubeDownscaler Metrics​

Here is a list of all the metrics currently exposed by the GoKubeDownscaler, divided into dry-run and production metrics. Common metrics are listed at the end, as they are always present.

GoKubeDownscaler Dry-Run Metrics​

  • metric_name: kubedownscaler_potential_downscaled_workloads

    • type: gauge
    • dimensions: namespace
    • description: Number of potential downscaled workloads managed by KubeDownscaler broken down by namespace.

  • metric_name: kubedownscaler_potential_upscaled_workloads

    • type: gauge
    • dimensions: namespace
    • description: Number of potential upscaled workloads managed by KubeDownscaler broken down by namespace.

  • metric_name: kubedownscaler_potential_excluded_workloads

    • type: gauge
    • dimensions: namespace
    • description: Number of potential workloads excluded from kubedownscaler management broken down by namespace.

  • metric_name: kubedownscaler_potential_current_saved_memory_bytes

    • type: gauge
    • dimensions: namespace
    • description: Number of potential bytes of memory saved by KubeDownscaler downscaling actions. Workloads managed by Keda ScaledObjects are not included in this metric.

  • metric_name: kubedownscaler_potential_current_saved_cpu_cores

    • type: gauge
    • dimensions: namespace
    • description: Number of potential cores of cpu saved by KubeDownscaler downscaling actions. Workloads managed by Keda ScaledObjects are not included in this metric.

GoKubeDownscaler Production Metrics​

  • metric_name: kubedownscaler_downscaled_workloads

    • type: gauge
    • dimensions: namespace
    • description: Number of downscaled workloads managed by KubeDownscaler broken down by namespace.

  • metric_name: kubedownscaler_upscaled_workloads

    • type: gauge
    • dimensions: namespace
    • description: Number of upscaled workloads managed by KubeDownscaler broken down by namespace.

  • metric_name: kubedownscaler_excluded_workloads

    • type: gauge
    • dimensions: namespace
    • description: Number of workloads excluded from KubeDownscaler management broken down by namespace.

  • metric_name: kubedownscaler_current_saved_memory_bytes

    • type: gauge
    • dimensions: namespace
    • description: Number of bytes of memory saved by KubeDownscaler downscaling actions. Workloads managed by Keda ScaledObjects are not included in this metric.

  • metric_name: kubedownscaler_current_saved_cpu_cores

    • type: gauge
    • dimensions: namespace
    • description: Number of cores of cpu saved by KubeDownscaler downscaling actions. Workloads managed by Keda ScaledObjects are not included in this metric.

GoKubeDownscaler Common Metrics​

  • metric_name: kubedownscaler_scaling_errors

    • type: gauge
    • dimensions: namespace, type
    • description: Number of scaling errors encountered during the scale process.

  • metric_name: kubedownscaler_cycle_duration_seconds

    • type: gauge
    • description: Duration of kubedownscaler cycle in seconds.

  • metric_name: kubedownscaler_cycle_executions_total

    • type: counter
    • description: Number of cycles completed by KubeDownscaler since being instantiated.
tip

When kubedownscaler_cycle_duration_seconds has a high value, it could be useful to review the resource requests and limits of the downscaler to ensure it has enough resources to operate efficiently. Another option is to enable concurrency If these options aren’t viable, we recommend setting the --max-retries-on-conflict flag This helps handle cases where the resource is modified by another component between the time it's fetched and updated by the downscaler.

List of GoKubeDownscaler Webhook Metrics​

Here is a list of all the metrics currently exposed by the GoKubeDownscaler Webhook, divided into dry-run and production metrics. Common metrics are listed at the end, as they are always present.

Webhook Dry-Run Metrics​

  • metric_name: kubedownscaler_potential_validate_workload_admission_requests_total
    • type: counter
    • dimensions: namespace, result
    • description: Number of workloads intercepted by kubedownscaler webhook broken down by namespace and result of mutating review.

Webhook Production Metrics​

  • metric_name: kubedownscaler_validate_workload_admission_requests_total
    • type: counter
    • dimensions: namespace, result
    • description: Number of workloads intercepted by kubedownscaler webhook broken down by namespace and result of mutating review.

Webhook Common Metrics​

  • metric_name: controller_runtime_webhook_latency_seconds

    • type: histogram
    • dimensions: webhook
    • description: Measures the latency of processing admission webhook requests broken down by webhook path.

  • metric_name: controller_runtime_webhook_requests_in_flight

    • type: counter
    • dimensions: webhook
    • description: Measures the webhook requests currently being processed broken down by webhook path.

  • metric_name: controller_runtime_webhook_requests_total

    • type: counter
    • dimensions: code, webhook
    • description: Measures the total number of webhook requests received broken down by http code webhook path.