End-to-End Automation Guide

This page is the English entry point for automated UI and Kubernetes acceptance. The detailed commands, workload examples, browser flow, and failure diagnosis are in the Chinese E2E guide.

What the end-to-end run checks

The run must exercise the real Java profiling path, not only load a healthy page:

• Kubernetes metadata accepts the target workload;
• CPU, allocation, and lock-delay profiles contain data;
• ClickHouse contains profile rows;
• ingestion reports accepted batches and any rejection, retry, drop, or truncation;
• the browser can complete the analysis workflow;
• the target workload has no unexpected additional restart.

For changes touching collector profiling, ingestion, ClickHouse, deployment, the demo workload, or the profile UI, follow the Real Profiling Acceptance Standard and Contributing.

Test layers

• UI smoke tests validate page loading and basic interactions.
• The real acceptance script validates collector, backend, ClickHouse, Web, and the JDK 17 workload together.
• Strict profiling acceptance requires target status, profile rows, ingestion evidence, browser evidence, bounded retention, and the restart invariant.

UI smoke tests alone cannot prove async-profiler attach or ClickHouse ingestion.

Detailed guide

Use the Chinese E2E automation guide for the current commands, Kubernetes examples, Playwright workflow, artifact layout, and failure matrix.