Search…
Built-in Metrics

Kubernetes Related Metrics

Kubernetes Pod Restarts

Every release policy must include a condition for Kubernetes Pod restarts.
Kubernetes Pod restarts are a good indicator of the basic performance of your service.
Kubernetes restarts a Pod if it fails its regular liveness probe. Kubernetes uses liveness probes to catch situations where an application is running but unable to make progress. Restarting a container in such a state can help to make the application more available, even if there are bugs.
Typically, a newly deployed version of a service will experience no restarts during a release, so you can set a low budget.
For example: fewer than two restarts per five-minute window.

Kubernetes Minimum Available Pods

We recommend including a condition for Kubernetes Minimum Available Pods in the first step of a multi-step release policy.
The Kubernetes Minimum Available Pods metric is a good indicator of whether a newly deployed version of your service is stable enough to be released.
This metric corresponds to the generic Kubernetes Does not have minimum availability message that is seen when a workload is deployed.
This can be caused by the nodes in the cluster lacking sufficient resources to run some or all of the requested Pods. It can also be caused by a misconfiguration, for example, a missing Kubernetes Secret.

Available Kubernetes Pods

The Available Kubernetes Pods metric can be a useful indicator of both the stability and the efficiency of your service.
Kubernetes marks a Pod as unavailable if it fails its regular readiness probe.
If you use a Horizontal Pod Autoscaler (HPA) to manage the number of Pods, a larger than expected number of available Pods can be an indicator that your new version is using more CPU or memory than expected.

REST and GraphQL API Metrics

API Success Rate

This metric is not available for the NGINX Ingress controller.
There are unresolved bugs in the NGINX Ingress controller which prevent access to the data needed for this metric. #4026 #3645
If your service exposes a REST API or GraphQL API, the API Success Rate metric indicates how many of the requests returned an error instead of the expected response.

API Request Duration

This metric is not available for the NGINX Ingress controller.
There are unresolved bugs in the NGINX Ingress controller that prevent access to the data needed for this metric. #4026 #3645
If your service exposes a REST API or GraphQL API, the API Request Duration metric indicates the average response time in milliseconds (ms).