Description

This is the first part of part of #728

Where available we'll enable native Prometheus endpoints for all our products.

Value

Adding the native endpoints now will allow all users to migrate to the new metrics endpoint (which might change labels) in their own time as we'll keep the JMX Exporter around for a one or more releases before we remove it.

Tasks

The JMX exporter is bundled with the following products and we need to check each of them if they offer a native alternative:

• enable built-in Prometheus servlet hdfs-operator#694
• chore: remove 2.4 support (including JMX exporter) hbase-operator#672
• hive: doesn't seem to provide built-in Prometheus metrics at this point.
• kafka: Doesn't seem to provide built-in Prometheus metrics, Grafana docs mention another standalone exporter which provides different metrics than the JMX exporter, but is out of scope of this issue since it's no native support.
• omid: no built-in prometheus support.
• spark : replace jmx with prometheus Replace the JMX exporter with native product metrics endpoints spark-k8s-operator#582
• spark: add jmx back fix: re-enable the jmx exporter spark-k8s-operator#585
• chore: add test and docs for metrics trino-operator#764
• Add native product metrics endpoints zookeeper-operator#954

For each of these products:

• Enable the native alternative where possible
• Enable the operators to use them where applicable
• Move any tests to use the new metrics
• Move any demos to use the new metrics

Additional tasks:

• Mark the "old" metrics as deprecated in the release notes
• Add an item to our deprecation tracker to remove the JMX Exporter later

Acceptance Criteria

• Operators expose metrics ports directly from products that support them.
• Updated tests where needed
• Updated product documentation
• Mark this as breaking change in the operator changelog if the exposed metrics change (very likely I'd say)