英文:
Using net.http.HttpClient in Tomcat causes memory leak
问题
我在一个基于Servlet的Web应用中使用Java新版本(自版本11以来)的HttpClient
,一切都运行得很顺利,除了在Tomcat重新启动Web应用时出现以下警告:
14-Aug-2020 09:21:16.996 WARNING [http-nio-8080-exec-18] org.apache.catalina.loader.WebappClassLoaderBase.clearReferencesThreads The web application [MyApp] appears to have started a thread named [HttpClient-3-SelectorManager] but has failed to stop it. This is very likely to create a memory leak. Stack trace of thread:
java.base/sun.nio.ch.WindowsSelectorImpl$SubSelector.poll0(Native Method)
java.base/sun.nio.ch.WindowsSelectorImpl$SubSelector.poll(WindowsSelectorImpl.java:357)
java.base/sun.nio.ch.WindowsSelectorImpl.doSelect(WindowsSelectorImpl.java:182)
java.base/sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:124)
java.base/sun.nio.ch.SelectorImpl.select(SelectorImpl.java:136)
java.net.http/jdk.internal.net.http.HttpClientImpl$SelectorManager.run(HttpClientImpl.java:867)
如何防止这个问题?我尝试使用自定义的ThreadFactory
仅返回守护线程:
HttpClient.newBuilder()
.executor(Executors.newSingleThreadExecutor((Runnable r) -> {
Thread t = new Thread(r);
t.setDaemon(true);
return t;
}))
.connectTimeout(Duration.ofSeconds(5))
.followRedirects(HttpClient.Redirect.NORMAL).build();
但警告仍然存在。
我正在使用OpenJDK 11.0.7和Tomcat 9。
如果还有其他问题,请告诉我。
英文:
I'm using the Java's new (since version 11) HttpClient
within a Servlet based web application:
private static final HttpClient HTTP_CLIENT =
.connectTimeout(Duration.ofSeconds(5))
.followRedirects(HttpClient.Redirect.NORMAL)
.build();
...
public void httpPostAsyncToEndpoint(WebEndpoint endpoint, Map<String,String> params) {
HttpRequest req = buildRequest(endpoint, params);
CompletableFuture<HttpResponse<String>> future = HTTP_CLIENT.sendAsync(req, HttpResponse.BodyHandlers.ofString());
future.thenAccept((HttpResponse<String> res) -> {
if (res.statusCode() >= 400) {
if (LOGGER.isErrorEnabled()) {
LOGGER.error("{} HTTP response returned from endpoint {}", endpoint, res.statusCode());
}
}
}).exceptionally(ex -> {
if (LOGGER.isErrorEnabled()) {
LOGGER.error("Could not audit event using endpoint {}", endpoint, ex);
}
return null;
});
}
Everything is working great, except that when the web application is restarted on Tomcat the following warning is produced:
14-Aug-2020 09:21:16.996 WARNING [http-nio-8080-exec-18] org.apache.catalina.loader.WebappClassLoaderBase.clearReferencesThreads The web application [MyApp] appears to have started a thread named [HttpClient-3-SelectorManager] but has failed to stop it. This is very likely to create a memory leak. Stack trace of thread:
java.base/sun.nio.ch.WindowsSelectorImpl$SubSelector.poll0(Native Method)
java.base/sun.nio.ch.WindowsSelectorImpl$SubSelector.poll(WindowsSelectorImpl.java:357)
java.base/sun.nio.ch.WindowsSelectorImpl.doSelect(WindowsSelectorImpl.java:182)
java.base/sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:124)
java.base/sun.nio.ch.SelectorImpl.select(SelectorImpl.java:136)
java.net.http/jdk.internal.net.http.HttpClientImpl$SelectorManager.run(HttpClientImpl.java:867)
How can I prevent this? I've attempted to use a custom ThreadFactory
which returns only daemon threads:
HttpClient.newBuilder()
.executor(Executors.newSingleThreadExecutor((Runnable r) -> {
Thread t = new Thread(r);
t.setDaemon(true);
return t;
}))
.connectTimeout(Duration.ofSeconds(5))
.followRedirects(HttpClient.Redirect.NORMAL).build();
but the warning persists.
I'm using OpenJDK 11.0.7 on Tomcat 9.
答案1
得分: 3
选择器管理线程将保持活动状态,只要对 HttpClient
的引用存在 - 或者只要客户端发起的操作仍在进行中。线程可能需要几秒钟的时间来检测到 HttpClient
不再被引用。因此,我认为您拥有的情况不是真正的泄漏 - 除非保存静态引用到 HttpClient
的类出于其他原因而在内存中保持固定。
英文:
The selector manager thread will remain alive as long as the reference to the HttpClient
is alive - or as long as an operation initiated by the client is still in progress. It may take a couple of seconds for the thread to detect that the HttpClient
is no longer referenced. So I don't believe that what you have is an actual leak - unless the class that held the static reference to the HttpClient
stays pinned in memory for other reasons.
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库,让每个人都能够通过互相帮助和分享经验来进步。
评论