java-concurrency-review

Check shared mutable state for check-then-act race conditions, verify volatile on fields read in loops or used in double-checked locking, and confirm consistent lock ordering to prevent deadlocks. Use ConcurrentHashMap atomic methods like computeIfAbsent instead of manual contains-then-put sequences.

@Async runs synchronously when called from within the same class because the self-invocation bypasses the Spring proxy. It also fails silently without @EnableAsync, on non-public methods, or when the method is called before the proxy is initialized.

Use virtual threads for I/O-bound tasks like HTTP calls, database queries, and file operations where you have thousands of concurrent tasks. They provide no benefit for CPU-bound work, which should use platform threads or ForkJoinPool instead.

Attach exceptionally() to recover with a fallback value, or handle() to process both success and failure outcomes. Without these, exceptions from supplyAsync are silently swallowed. Java 9+ also adds orTimeout() and completeOnTimeout() for deadline handling.

No, SecurityContextHolder is ThreadLocal-bound, so the context is null inside async threads by default. Wrap your executor with DelegatingSecurityContextAsyncTaskExecutor to propagate the security context to async tasks.

Nested compute calls on the same ConcurrentHashMap can deadlock because the bin is locked during computation. Compound operations like contains-then-put are also non-atomic; use putIfAbsent or computeIfAbsent for atomicity instead.