DISCLAIMER This component is part of the rating-operator application and is not designed to work as a standalone program.
Rating-operator-engine uses custom resources to configure its workers and trigger rating workloads, with a variable timeframe. Each workers holds their own configuration, that can be updated at runtime, and query, rate and store data in PostgreSQL through the rating-operator-api.
The configuration, named RatingRuleInstances, can be sumed up as follow:
- Name
The name of the rated metric. It is the name by which you'll be able to query your metric afterward.
- Timeframe (Base value of 60s)
This parameters control the time between each data query, in seconds. Depending on the precision or reactivity need of a given metric, you can adapt this parameter to suit your need. Considering Prometheus expose datapoints for each seconds, "timeframe" imply the size of the metric set you receive. By using, for example, 3600s as the timeframe, you will receive, every hour, 3600 datapoints that will then be sumed into one.
- Metric
A prometheus query expression (promQL) to be executed every "timeframe". It can be enhanced with variables, defined in the RatingRules configuration, and made available through Prometheus. More information on RatingRules can be read in the rating-operator (Custom Resources) document.
Below an example of a RatingRuleInstance:
apiVersion: rating.smile.fr/v1
kind: RatingRuleInstances
metadata:
name: my-test-rule # The name of the RatingRuleInstances object
spec:
metric: sum(kube_pod_container_resource_requests_cpu_cores) by (pod, namespace, node) # The prometheus query to execute
name: pods_usage_cpu # The name of the metric
timeframe: 60s # The time between each queryFor more informations on how to configure and use the component, please refer to the rating-operator documentation.