WHAT

The new metrics introduced by the work of this MGDSTRM-10148 are required to have the instance_name and instance_id labels. This change will help bring them into being.

WHY

Supports the use-case being delivered by this Epic. This also takes a step towards MGDSTRM-7080.

HOW

The metrics produced by fleetshard already have the instance_name label. We need to add the instance_id.

• Change fleetshard MetricManager to populate a instance_id label from org.bf2.operator.resources.v1alpha1.ManagedKafka#ID managed kafka resource label. This will add the label onto all instance specific metrics emitted by fleeshard, including kafka_instance_connection_limit desired by this change.

In RHOSAK the kafka metrics already have an instance_name label. This is produced by this relabelling rule which takes the strimzi_io_cluster label produced by Strimzi and maps it into instance_name. This is no label containing the instance's id We make a change to cause the scrape to emit a bf2_org_id label containing the instance's id.

• Change fleetshard org.bf2.operator.operands.KafkaCluster#buildKafkaLabels to include the instance's id as a label bf2.org/id on the Kakfa CR.

Note that:

• Strimzi will propagate the label to the pods.
• Prometheus automatic creates metric labels from the pod's labels during ingestion. This will mean that metrics such as kafka_server_socket_server_metrics_connection_count will automatically have a bf2_org_id label.
• MGDSTRM-10287 will use a relabelled rule to map bf2_org_id to instance_id.

We should verify that the addition of these labels won't upset existing alerts or dashboards.

DONE

• Changes made to fleetshard with appropriate unit tests.