Spring Boot应用中监控tomcat,druid连接池状态

约定不等于承诺〃 2022-05-25 10:16 1073阅读 0赞

这篇文章记录一下如何在spring boot中监控tomcat,druid连接池状态，之前我写过一篇文章介绍过类似的，具体背景和监控数据的收集部分这里就不赘述，可以参考[在SPRING BOOT应用监控线程池的状态][SPRING BOOT]

## 背景 ##

随着流量的增大，我们急需监控各个微服务部署的tomcat和数据库连接池状态，以此来了解线上连接池配置是否满足要求，不存在性能上的问题。

## 数据收集 ##

### JMX ###

再介绍怎么监控连接池状态之前，先来了解一下JMX, JMX在Java编程语言中定义了应用程序以及网络管理和监控的体系结构、设计模式、应用程序接口以及服务。通常使用JMX来**监控系统的运行状态或管理系统的某些方面**，比如清空缓存、重新加载配置文件等。

tomcat和druid在启动后都会在JMX中注册自己的Mbean, 其中Mbean有关于连接池状态的属性，因此只需要定时读取Mbean中的这部分属性就能做到对线程池的监控。

怎么试着查看JMX呢？最简单的办法就是通过jvisualvm，打开已经启动的spring boot项目，当然前提是要在jvisualvm中提前安装好mbeans的插件，之后我们能发现有一个Mbeans标签，点击它就能看到所有注册到JMX的Mbean， 其中就有Tomcat和Druid的Mbean，顺着它们的树状结构查看它们支持的所有属性值，这些属性值就是正在运行的spring boot程序的当前状态，挑出我们最关心的属性继续往下看。

还是和之前一样利用spring boot actuator, 它能和statsd集成，并定时将/metrics中的监控数据发送出去，因此我们只需要实现PublicMetrics接口，而在metrics方法中读取tomcat和druid的Mbean数据，具体可以参考下面的代码：

### 监控Tomcat连接池 ###

@Component
    public class TomcatPublishMetrics implements PublicMetrics {
    
        public static final Logger LOG = LoggerFactory.getLogger(TomcatPublishMetrics.class);
    
        @Autowired
        private MBeanServer mbeanServer;
    
        @Override
        public Collection<Metric<?>> metrics() {
            try {
                Set<ObjectName> objectNames = mbeanServer.queryNames(new ObjectName("Tomcat:type=ThreadPool,*"), null);
                Collection<Metric<?>> result = new ArrayList<>();
                if (objectNames != null && objectNames.size() == 1) {
                    for (ObjectName objectName : objectNames) {
                        int maxThreads = (Integer) mbeanServer.getAttribute(objectName, "maxThreads");
                        int currentThreadCount = (Integer) mbeanServer.getAttribute(objectName, "currentThreadCount");
                        int currentThreadsBusy = (Integer) mbeanServer.getAttribute(objectName, "currentThreadsBusy");
                        long connectionCount = (Long) mbeanServer.getAttribute(objectName, "connectionCount");
                        int maxConnections = (Integer) mbeanServer.getAttribute(objectName, "maxConnections");
                        result.add(new Metric<Number>("tomcat.thread.busy", currentThreadsBusy, new Date()));
                        result.add(new Metric<Number>("tomcat.thread.max", maxThreads, new Date()));
                        result.add(new Metric<Number>("tomcat.thread.current", currentThreadCount, new Date()));
                        result.add(new Metric<Number>("tomcat.thread.connectionCount", connectionCount, new Date()));
                        result.add(new Metric<Number>("tomcat.thread.maxConnections", maxConnections, new Date()));
                    }
                } else {
                    LOG.warn("Tomcat thread pool monitor failed");
                }
                return result;
            } catch (Exception e) {
                LOG.error("Tomcat thread pool monitor exception", e);
            }
            return Collections.emptyList();
        }
    }

### 监控Druid连接池 ###

@Component
    public class DruidPoolPublishMetrics implements PublicMetrics {
    
        public static final Logger LOG = LoggerFactory.getLogger(DruidPoolPublishMetrics.class);
    
        @Autowired
        private MBeanServer mbeanServer;
    
        @Override
        public Collection<Metric<?>> metrics() {
            try {
                Set<ObjectName> objectNames = mbeanServer.queryNames(new ObjectName("com.alibaba.druid.pool:type=DruidDataSource,*"), null);
                Collection<Metric<?>> result = new ArrayList<>();
                if (objectNames != null && objectNames.size() == 1) {
                    for (ObjectName objectName : objectNames) {
                        long errorCount = (Long) mbeanServer.getAttribute(objectName, "ErrorCount");
                        int waitThreadCount = (Integer) mbeanServer.getAttribute(objectName, "WaitThreadCount");
                        long resetCount = (Long) mbeanServer.getAttribute(objectName, "ResetCount");
                        int activeCount = (Integer) mbeanServer.getAttribute(objectName, "ActiveCount");
                        long rollbackCount = (Long) mbeanServer.getAttribute(objectName, "RollbackCount");
                        int maxActive = (Integer) mbeanServer.getAttribute(objectName, "MaxActive");
                        result.add(new Metric<Number>("druid.pool.errorCount", errorCount, new Date()));
                        result.add(new Metric<Number>("druid.pool.waitThreadCount", waitThreadCount, new Date()));
                        result.add(new Metric<Number>("druid.pool.resetCount", resetCount, new Date()));
                        result.add(new Metric<Number>("druid.pool.activeCount", activeCount, new Date()));
                        result.add(new Metric<Number>("druid.pool.rollbackCount", rollbackCount, new Date()));
                        result.add(new Metric<Number>("druid.pool.maxActive", maxActive, new Date()));
                    }
                } else {
                    LOG.warn("DruidPool monitor failed");
                }
                return result;
            } catch (Exception e) {
                LOG.error("DruidPool monitor exception", e);
            }
            return Collections.emptyList();
        }
    }

## 总结 ##

这篇文章介绍了通过读取JMX中相应的Mbean的数据进行收集并监控，如果其他的监控指标注册在JMX中，例如对redis的连接池，也可以采用类似的方法收集。

欢迎关注我的个人的博客[www.zhijianliu.cn][], 虚心求教，有错误还请指正轻拍，谢谢

[SPRING BOOT]: http://www.zhijianliu.cn/2018/05/05/%E5%9C%A8spring-boot%E5%BA%94%E7%94%A8%E7%9B%91%E6%8E%A7%E7%BA%BF%E7%A8%8B%E6%B1%A0%E7%9A%84%E7%8A%B6%E6%80%81/
[www.zhijianliu.cn]: http://www.zhijianliu.cn/