Crossplane has two aspects to consider for monitoring and observability. Monitoring Crossplane itself, and composing capabilities to offer monitored resources upon issuing a claim.

Crossplane and provider observability

To enable metrics for Crossplane, you must configure the option metrics.enabled and set it to true. Do this during the initial install of Crossplane, or at any time while it is running. For instance, when using the Upbound up uxp install or up uxp upgrade command.

The metrics that Crossplane and its providers emit are available at port 8080 under the /metrics URL path. Pick up the metrics feed with Prometheus or a Prometheus compatible scraper. The specific metrics inform about the Crossplane and provider Kubernetes pod performance. They include the following categories:

• Certwatcher Read Requests and Errors
• Controller Runtime Reconciliation Counts
• Go Memory and Threads
• Leader Election Status
• Process CPU and Memory
• Rest Client Requests
• Work queue Depth and Processing Duration
• Provider Processes

After you’ve enabled metrics emission in Crossplane, you should use a tool to collect the emitted logs and metrics. Your org may already be using one of the third-party solutions that integrate with Kubernetes for monitoring. Examples include as Datadog, Grafana, or New Relic. You should use these with Crossplane.

Managed resource observability

When you design your custom APIs, they map to provider capabilities through compositions. It’s a good practice to identify which observability features to provision along with those resources. Amazon AWS, Google GCP and Microsoft Azure provider families all include aspects of observability provisioning features. For example, provider-gcp-monitoring gives you access to a rich set of Google Cloud metrics. Amazon provider-aws-cloudwatch allows setting of alarm triggers and Microsoft provider-azure-logz enables monitoring of event logs. Also, the Upbound Marketplace has a couple of related providers, such as provider-grafana and provider-newrelic.

You may want to automatically create dashboards for the resources you might compose when receiving a resource claim from your application teams. It creates a better experience for them because they now have insight into the health of the resources that they just claimed. Based on your organization’s needs and the specific requirements of your own customers, engage in discussions with them what they would like to see.