Do more with less: Unit and integration testing is broken
summary
Argues that modern observability can replace much test boilerplate: a production music service of ~150k lines and 20+ integrations runs ~300,000 transactions per hour with 15% test coverage while using distributed tracing and Instana dashboards to correlate traces, P50–P99 latencies, errors and infrastructure. Shows concrete fixes found and verified — a Hibernate O(n log n) query causing tail-latency, and an auth-token expiry solved with Redis caching — proving monitoring pinpoints failures and validates remedies in production.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 2:13Culture of testing critique
- 6:49What not to test
- 7:11Distributed tracing explained
- 10:03Single Music case study
- 13:16Production dashboards and observability
- 18:35Summary and Q&A
related talks
10mBeyond “try harder”: Effective strategies to tackle bugs
30mReducing Chaos in Production: Testing vs. Monitoring
44mTesting the untestable - patterns and use cases analysis
01h 02mThe Art & Joy of Testing in Production
57mAdvanced Testing Techniques: Tips from the trenches
25mWriting Testable Code
28mProduction Testing through Monitoring
23mA peek into observability from a tester's lens