spring taskExecutor redis redisMessageListenerContailner jvm 线程优化

最新推荐文章于 2025-03-24 10:19:39 发布

javaee_ssh

最新推荐文章于 2025-03-24 10:19:39 发布

阅读量4.5k

点赞数 2

文章标签： redis spring-session spring-data-redis

本文链接：https://blog.youkuaiyun.com/javaee_ssh/article/details/76896132

版权

本文介绍了一位测试工程师在进行系统压力测试时遇到的问题：由于Redis消息监听容器(redisMessageListenerContainer)未合理配置线程池，导致系统进程中产生了大量线程并最终被杀掉。文章详细分析了redisMessageListenerContainer的工作原理，并给出了通过配置自定义线程池来解决问题的具体方案。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

背景：

测试工程师给系统做压力测试，发现JVM产生了大量的redisMessageListenerContainer线程,系统跑着跑着进程就被kill掉了

查看源码：

进入redisMessageListenerContainer 类打断点进去，查看内部代码

redisMessageListenerContainer.java

public void onMessage(Message message, byte[] pattern) {
		Collection<MessageListener> listeners = null;
		// if it's a pattern, disregard channel
		if (pattern != null && pattern.length > 0) {
			listeners = patternMapping.get(new ByteArrayWrapper(pattern));
		} else {
			pattern = null;
			// do channel matching first
			listeners = channelMapping.get(new ByteArrayWrapper(message.getChannel()));
		}
		if (!CollectionUtils.isEmpty(listeners)) {
			dispatchMessage(listeners, message, pattern);
		}
	}

}

	private void dispatchMessage(Collection<MessageListener> listeners, final Message message, final byte[] pattern) {
		final byte[] source = (pattern != null ? pattern.clone() : message.getChannel());
		for (final MessageListener messageListener : listeners) {
			taskExecutor.execute(new Runnable() {// 使用了线程池

				public void run() {
					processMessage(messageListener, message, source);
				}
			});
		}
	}

	// 测试工程师给系统做压力测试，发现JVM产生了大量的redisMessageListenerContainer线程,系统跑着跑着进程就被kill掉了
	public void afterPropertiesSet() {
		if (taskExecutor == null) {// 如果没有设置默认的使用的是SimpleAsyncaskExecutor,这是一个简单的执行器，并不有线程池功能
			manageExecutor = true;
			taskExecutor = createDefaultTaskExecutor();
		}
		if (subscriptionExecutor == null) {
			subscriptionExecutor = taskExecutor;
		}
		initialized = true;
	}

	protected TaskExecutor createDefaultTaskExecutor() {
		String threadNamePrefix = (beanName != null ? beanName + "-" : DEFAULT_THREAD_NAME_PREFIX);
		return new SimpleAsyncTaskExecutor(threadNamePrefix);
	}

RedisHttpSessionConfiguration.Java

    @Bean
    public RedisMessageListenerContainer redisMessageListenerContainer(RedisConnectionFactory connectionFactory, RedisOperationsSessionRepository messageListener) {
        RedisMessageListenerContainer container = new RedisMessageListenerContainer();
        container.setConnectionFactory(connectionFactory);
        if(this.redisTaskExecutor != null) {
            container.setTaskExecutor(this.redisTaskExecutor);
        }

        if(this.redisSubscriptionExecutor != null) {
            container.setSubscriptionExecutor(this.redisSubscriptionExecutor);
        }

        container.addMessageListener(messageListener, Arrays.asList(new PatternTopic[]{new PatternTopic("__keyevent@*:del"), new PatternTopic("__keyevent@*:expired")}));
        container.addMessageListener(messageListener, Arrays.asList(new PatternTopic[]{new PatternTopic(messageListener.getSessionCreatedChannelPrefix() + "*")}));
        return container;
    }
    
    @Qualifier("springSessionRedisTaskExecutor")
    public void setRedisTaskExecutor(Executor redisTaskExecutor) {
        this.redisTaskExecutor = redisTaskExecutor;
    }

解决方案：

XML中添加一个redis线程池，问题解决

<bean id="springSessionRedisTaskExecutor" class="org.springframework.scheduling.concurrent.ThreadPoolTaskExecutor">
		<!-- 核心线程数 -->
		<property name="corePoolSize" value="10" />
		<!-- 最大线程数 -->
		<property name="maxPoolSize" value="300" />
		<!-- 队列最大长度 -->
		<property name="queueCapacity" value="500" />
		<!-- 线程池维护线程所允许的空闲时间，默认为60s -->
		<property name="keepAliveSeconds" value="60" />
</bean>