Kube Burner: Kubernetes Performance and Scale Testing

Introduction

Performance regressions in Kubernetes are rarely obvious in small test clusters. They show up when you hit “real” scale: many namespaces, lots of objects, frequent deploys, noisy neighbors, and dashboards that start lagging right when you need them.

Kube Burner, accepted as a CNCF Sandbox project in 2024, is a performance and scale test orchestration toolset for Kubernetes. It focuses on creating, deleting, and patching resources at scale—paired with Prometheus metric collection and alerting—so you can reproduce pressure patterns and compare environments and changes over time.

Scale Testing

• Resource creation enables creating large numbers of Kubernetes resources.
• Resource deletion enables deleting resources at scale.
• Resource patching enables patching resources at scale.
• Workload generation enables generating realistic workloads for testing.

Performance Metrics

1. Prometheus integration enables collection of performance metrics.
2. Metric collection provides comprehensive metrics for performance analysis.
3. Alerting capabilities enable alerting on performance thresholds.
4. Reporting provides detailed performance reports.

Test Orchestration

• Test execution enables orchestration of complex test scenarios.
• Workload management provides management of test workloads.
• Resource management enables management of test resources.
• Cleanup automation enables automatic cleanup of test resources.

Use Cases

• Performance testing enables testing Kubernetes cluster performance.
• Scale testing enables testing cluster scalability.
• Load testing enables testing cluster under load.
• Capacity planning enables planning cluster capacity.

Practical notes (making scale tests actionable)

• Test like you operate: mirror your real object mix and churn patterns (deploy frequency, autoscaling events, CRDs), not just raw pod counts.
• Pick a baseline and keep it: scale tests are most useful when you can compare apples-to-apples across cluster versions, config changes, and node types.
• Include failure signals: watch not only “throughput” but also API latencies, controller backlogs, and error budgets—these often tell the story first.

Summary

Kube Burner represents a significant advancement in Kubernetes performance testing, providing teams with powerful capabilities for testing cluster performance and scalability.