我们的订单服务有一个 UserContext保存当前登录使用者的信息，在 Controller 层设置。在 Service 层读取：
public class UserContext {
private static final ThreadLocal CONTEXT = new ThreadLocal<>;private Long userId;说起来，private String username;private List permissions;// 权限列表，比较大
public static void set { CONTEXT.set;}
public static UserContext get { return CONTEXT.get;}
public static void clear { CONTEXT.remove;}
}
在拦截器里设置和清理：
@Component
public class UserContextInterceptor implements HandlerInterceptor {
@Override
public boolean preHandle(HttpServletRequest request。HttpServletResponse response,Object handler) {
// 从 token 解析使用者信息，设置到 ThreadLocal
UserContext ctx = parseToken);UserContext.set;老实说，return true;}
@Override
public void afterCompletion(HttpServletRequest request。HttpServletResponse response,Object handler,Exception ex) {
UserContext.clear;// 请求结束，清理 ThreadLocal
}
}
This leak is **gradual** – 同一 key 被新值覆盖。
但如果 项目中出现多个不同 key，它们各自独立泄 漏，不相互影响。
四、为什么 ThreadLocal 会泄漏？底层原理
ThreadLocal 的存储结构
Per‑thread storage is a custom hash map called ThreadLoc alMap . Its key is a weak reference to ThreadLoc al ,and its value is a strong reference to whatever you stored:
executor.submit -> {
try {
UserCon tex t.set;不过，// 设置
doAsyncWork;} finally {
Us erCon tex t.clear;// 必须清理，}
});This is baseline.
至于方案二，自定义 ThreadPoolTaskExecutor 在层面统一清理
If you sprinkle try‑finally everywhere。someone will inevitably forget.
/@Component
public class T hreadL ocalAwareTaskExecutor extends T hreadPoo lTaskExecutor {
@Override
public void execute {
// Capture snapshot of submit‑thread's TLs
Map snapshot = captureThreadLocals;super.execute -> {
try {
restoreThreadLocals;话说回来，command.run;} finally {
cleanAllThreadLocals;}
})，}
private Map captureT hreadLocals {
// Use TTL's snapshot ability or maintain your own registry
Map snap = new HashMap<>;// ，populate snap…return snap，}
private void restoreT hreadLocals {
snap.forEach -> {
if tl.set;else tl.remove;})，}
private void cleanAllT hreadLocals {
UserCon tex t.clear;其实，TraceCo ntext.clear;// ，or TLs…}
}
方案三这方面，使用 TransmittableThreadLocal
The open‑source library from Alibaba does exactly what you need.
/ / 把 ThreadLo cal 换成 TransmittableThre adLo cal
public class U serCo ntext {
private static final TransmittableThrea dLocal
public static void set{ CON TEX T.se t;}
public static Us erCon tex t get{ retu rn CONT EXT.g et;}
public static vo id clear{ CON TEX T.re mo ve;说起来，}
}
// Spring Boot bean example
@Bean
public T hreadPo olTaskExecu tor taskExecutor {
T hreadPo olTaskExecu tor exec = new T hreadPo olTaskExecu tor;exec.setCorePoolSize;exec.setMaxPoolSize;exec.setTa skDecorator;// ★关键★ TTL 包装
exec.initialize;return exec，怎么说呢，}
