What problem does it solve? Running Apache Kafka on Kubernetes through the Strimzi operator involves dozens of interlocking decisions — KRaft node pool topology, broker resource limits, Cruise Control capacity contracts, listener selection, storage classes, and JVM tuning — where a single mismatch silently corrupts performance-test results or destabilizes the cluster. This Skill encodes twelve non-negotiable rules and a saturation-band sizing methodology so clusters are perf-testable, sizing-defensible, and upgrade-boring. ## Core Features & Use Cases - KRaft-only cluster authoring: Declarative Kafka and KafkaNodePool resources with disjoint controller and broker pools, hard pod anti-affinity, and rack awareness. - Capacity contract enforcement: The exact BrokerCapacity API schema (cpu, inboundNetwork, outboundNetwork regex patterns, per-broker overrides) mirrored 1:1 against KafkaNodePool.spec.resources.limits so Cruise Control rebalances reflect real headroom. - Performance-testing playbook: Listener selection for synthetic load (internal plaintext for headline runs, TLS/SCRAM measured separately), perf-grade storage classes (Premium SSD v2, gp3, hyperdisk), mandatory Grafana panels before firing k6, and a Bronze-to-Platinum+ tier ladder across Kafka's four limiting resources (CPU, network-in, network-out, log disk). - Use Case: Before running an xk6-kafka load test against a hex-scaffold consumer/producer stack, use this Skill to author the cluster manifests, verify quotas are off, confirm metrics are wired, and rank the first saturated resource against established PostgreSQL baselines. ## Quick Start Ask the AI to author a perf-grade Strimzi Kafka cluster with split controller and broker node pools, Cruise Control capacity mirroring, and a k6 load-test plan for your Kubernetes namespace.