backend-java

Use scenario-based prompts grounded in production incidents, such as a container OOMKilled despite the heap never filling, rather than asking candidates to recite JVM memory regions. Pair every concept question with a follow-up asking how the issue manifests in production and how to verify it.

Core topics include JVM memory and OOM diagnosis, GC selection and tuning, thread pool design, locks and JUC tools, the Java Memory Model, Spring IoC/AOP failure cases, MyBatis/JPA with connection pools, microservice components, distributed locks, and performance troubleshooting with Arthas or profilers.

Campus hiring focuses on fundamentals like HashMap internals, synchronized versus ReentrantLock, JVM memory regions, and Bean lifecycle. Experienced hiring emphasizes production diagnosis such as frequent Full GC, thread pool sizing, @Transactional failure scenarios, and distributed lock pitfalls.

Only ask about JDK 21 virtual threads when the candidate's project actually uses newer JDK versions. For projects still on JDK 8, skip virtual threads and ZGC, but always expect candidates to read GC logs and thread dumps regardless of version.

Memorization questions like listing GC algorithms fail to distinguish candidates who have actually diagnosed production issues. Major companies now prefer presenting a code snippet or symptom and asking the candidate to identify the problem, revealing real troubleshooting experience.