Java Developer Interview Questions and Answers

Questions and model answers

9 questions you are likely to hear.

1. 1

   How does HashMap work internally?

   Why they ask: This is the most asked core Java question and tests understanding of hashing, collisions and performance.

   Sample answer: HashMap stores entries in an array of buckets. When I put a key, it calls hashCode, spreads the bits and uses the result to pick a bucket index. If two keys land in the same bucket, entries form a linked list, and since Java 8, once a bucket passes eight entries and the table is large enough, it converts to a red-black tree so lookups stay O(log n). When size exceeds capacity times the load factor of 0.75, the table doubles and entries are redistributed. That is why mutable keys are dangerous: if hashCode changes after insertion, the entry becomes unreachable.

2. 2

   Why must you override hashCode when you override equals?

   Why they ask: Interviewers test the contract between the two methods and practical consequences.

   Sample answer: The contract says equal objects must have equal hash codes. Hash-based collections first use hashCode to find a bucket and then equals to find the exact entry. If I override equals but not hashCode, two logically equal objects may land in different buckets, so a HashSet can hold duplicates and a HashMap get can return null for a key that is present. I usually generate both from the same fields, or use a Java record, which provides both automatically. I once fixed a deduplication bug in a customer import job caused exactly by this.

3. 3

   What is the difference between synchronized, ReentrantLock and ConcurrentHashMap?

   Why they ask: This checks practical concurrency knowledge.

   Sample answer: synchronized is a built-in monitor lock that is simple and released automatically when the block exits. ReentrantLock gives more control: tryLock with a timeout, interruptible locking and optional fairness, but I must unlock in a finally block. ConcurrentHashMap is a thread-safe map that avoids locking the whole map, using CAS operations and fine-grained locking per bin, and offers atomic methods like computeIfAbsent. In a pricing service, replacing a synchronized HashMap cache with ConcurrentHashMap and computeIfAbsent removed a lock contention hotspot and improved throughput noticeably under load tests.

4. 4

   How would you find and fix a memory leak in a Java service?

   Why they ask: This tests production debugging skills, which matter more than theory for experienced roles.

   Sample answer: I start with monitoring: if heap usage after each full GC keeps climbing, that suggests a leak. I take a heap dump with jcmd or enable HeapDumpOnOutOfMemoryError, then analyse it in Eclipse MAT to find the dominator tree and the objects retaining the most memory. Common causes are static collections that only grow, caches without eviction, unclosed resources and ThreadLocal values in thread pools. In one Spring Boot service, MAT showed a static map of session objects that was never cleaned. Switching to a Caffeine cache with expiry fixed weekly out-of-memory restarts.

5. 5

   Explain how @Transactional works in Spring and a common pitfall.

   Why they ask: Transactions are core to most enterprise Java work and often misunderstood.

   Sample answer: @Transactional works through a Spring proxy: when a call enters the proxy, it starts or joins a transaction, and on return it commits, or rolls back on unchecked exceptions by default. The most common pitfall is self-invocation: calling a transactional method from another method in the same class bypasses the proxy, so no transaction starts. Another is that checked exceptions do not trigger rollback unless I set rollbackFor. I hit the self-invocation issue in an order service where partial writes stayed in the database. Moving the method into a separate service bean fixed it.

6. 6

   What is the N+1 query problem in Hibernate and how do you solve it?

   Why they ask: This is a very common performance issue in JPA applications.

   Sample answer: N+1 happens when I load a list of parent entities with one query and Hibernate then runs one extra query per parent to load a lazy association. Loading 200 orders and their items can mean 201 queries. I detect it by enabling SQL logging or using a tool like Hibernate statistics. Fixes include a JOIN FETCH in JPQL, an entity graph, or batch fetching with a batch size setting. For read-only screens I often use a DTO projection instead. In one reporting API this reduced the queries per request from a few hundred to two, and response time fell from about 3 seconds to under 300 milliseconds.

7. 7

   How do you make microservices resilient when a downstream service is slow?

   Why they ask: Interviewers check practical distributed systems knowledge.

   Sample answer: I set explicit timeouts on every outbound call, because default timeouts are often far too long. I add a circuit breaker with Resilience4j so that after a failure threshold, calls fail fast and return a fallback, like cached data or a clear error. Retries are limited, use exponential backoff and are only applied to idempotent operations. Bulkheads keep one slow dependency from exhausting all threads. In a payments platform, adding a 2-second timeout and a circuit breaker on a fraud-check service stopped a slowdown in that service from taking down the entire checkout flow.

8. 8

   What are virtual threads in Java 21 and when would you use them?

   Why they ask: This checks whether you keep up with modern Java.

   Sample answer: Virtual threads are lightweight threads managed by the JVM rather than the operating system, so I can run hundreds of thousands of them. When a virtual thread blocks on I/O, the JVM unmounts it from the carrier thread, which lets simple blocking code scale like asynchronous code. They suit I/O-heavy workloads such as REST calls and database queries, not CPU-heavy tasks. Pitfalls include pinning when blocking inside synchronized blocks and libraries that rely on ThreadLocal heavily. In Spring Boot 3.2 and later I can enable them with a single property and then load test before rolling out.

9. 9

   Tell me about a difficult production issue you resolved.

   Why they ask: This behavioral-technical question reveals ownership and debugging approach.

   Sample answer: Our order service at a retail client began timing out every day around noon. Thread dumps showed most request threads waiting on database connections from HikariCP. Looking at slow query logs, a reporting query added in the last release was running a full table scan at peak time and holding connections for seconds. I added a composite index on the filtered columns, moved the report to a read replica and set a connection timeout with a clear alert. Timeouts stopped the same day, and I added slow query checks to our release checklist.

Keep going.

• Full Stack Developer Interview Questions and Answers
• Python Developer Interview Questions and Answers
• Java job support
• Coding Interview Tips for Developers
• System Design Interview Basics
• Coding interview help