[3.0] AbstractDirectory存在NPE以及connectivityCheckFuture重新赋值的问题

[zrlw]

Environment

• Dubbo version: 3.0 首先说一下NPE，如下所示 https://github.com/apache/dubbo/runs/4113012092?check_suite_focus=true

  2021-11-05T03:07:48.1927405Z [INFO] java.lang.NullPointerException
  2021-11-05T03:07:48.1933640Z [INFO]     at org.apache.dubbo.rpc.cluster.RouterChain.route(RouterChain.java:174)
  2021-11-05T03:07:48.1936223Z [INFO]     at org.apache.dubbo.registry.integration.DynamicDirectory.doList(DynamicDirectory.java:183)
  2021-11-05T03:07:48.1944252Z [INFO]     at org.apache.dubbo.rpc.cluster.directory.AbstractDirectory.list(AbstractDirectory.java:186)
  2021-11-05T03:07:48.1949775Z [INFO]     at org.apache.dubbo.rpc.cluster.support.AbstractClusterInvoker.list(AbstractClusterInvoker.java:392)
  2021-11-05T03:07:48.2016991Z [INFO]     at org.apache.dubbo.rpc.cluster.support.AbstractClusterInvoker.invoke(AbstractClusterInvoker.java:329)
  2021-11-05T03:07:48.2019606Z [INFO]     at org.apache.dubbo.monitor.support.MonitorFilter.invoke(MonitorFilter.java:92)
  2021-11-05T03:07:48.2022000Z [INFO]     at org.apache.dubbo.rpc.cluster.filter.FilterChainBuilder$FilterChainNode.invoke(FilterChainBuilder.java:84)
  2021-11-05T03:07:48.2024621Z [INFO]     at org.apache.dubbo.rpc.protocol.dubbo.filter.FutureFilter.invoke(FutureFilter.java:51)
  2021-11-05T03:07:48.2027274Z [INFO]     at org.apache.dubbo.rpc.cluster.filter.FilterChainBuilder$FilterChainNode.invoke(FilterChainBuilder.java:84)
  2021-11-05T03:07:48.2030322Z [INFO]     at org.apache.dubbo.rpc.cluster.filter.support.ConsumerContextFilter.invoke(ConsumerContextFilter.java:108)
  2021-11-05T03:07:48.2033091Z [INFO]     at org.apache.dubbo.rpc.cluster.filter.FilterChainBuilder$FilterChainNode.invoke(FilterChainBuilder.java:84)
  2021-11-05T03:07:48.2036348Z [INFO]     at org.apache.dubbo.rpc.cluster.support.wrapper.AbstractCluster$ClusterFilterInvoker.invoke(AbstractCluster.java:92)
  2021-11-05T03:07:48.2039691Z [INFO]     at org.apache.dubbo.rpc.cluster.support.wrapper.MockClusterInvoker.invoke(MockClusterInvoker.java:95)
  2021-11-05T03:07:48.2043182Z [INFO]     at org.apache.dubbo.registry.client.migration.MigrationInvoker.invoke(MigrationInvoker.java:276)
  2021-11-05T03:07:48.2046470Z [INFO]     at org.apache.dubbo.rpc.proxy.InvokerInvocationHandler.invoke(InvokerInvocationHandler.java:92)
  2021-11-05T03:07:48.2048634Z [INFO]     at org.apache.dubbo.common.bytecode.proxy5.hello(proxy5.java)
  2021-11-05T03:07:48.2051148Z [INFO]     at org.apache.dubbo.config.bootstrap.MultiInstanceTest.testAsyncExportAndReferServices(MultiInstanceTest.java:849)

此外，上面异常日志信息暴露出MultiInstanceTest存在的问题： provider启动完成并不意味着reference端收到了注册中心的服务实例通知，此时客户端直接调用会因找不到服务注册信息而失败。

[zrlw]

派生类和AbstractDirectory一起对AbstractDirectory的成员invokers，validInvokers进行操作违反了开闭原则，代码散落在各个类里面也不利于维护。

[zrlw]

再说一下this.connectivityCheckFuture被创建的新任务重新赋值的问题： checkConnectivity异步创建的任务在结束前判断是否有新task，如果有就再次调用checkConnectivity：

1. this.connectivityCheckFuture在checkConnectivity方法返回调用者之后又被赋了新值，但调用者并不知情；
2. 调用者在旧的connectivityCheckFuture上等待的结果就是新task未执行完，调用者就继续执行后面的代码，然后出问题，比如： https://github.com/apache/dubbo/runs/4130565263?check_suite_focus=true

   2021-11-07T11:46:56.1815610Z [ERROR] Tests run: 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.19 s <<< FAILURE! - in org.apache.dubbo.rpc.cluster.support.ConnectivityValidationTest
   2021-11-07T11:46:56.1818867Z [ERROR] org.apache.dubbo.rpc.cluster.support.ConnectivityValidationTest.testRetry  Time elapsed: 0.007 s  <<< FAILURE!
   2021-11-07T11:46:56.1820680Z org.opentest4j.AssertionFailedError: expected: <2> but was: <1>
   2021-11-07T11:46:56.1822662Z    at org.apache.dubbo.rpc.cluster.support.ConnectivityValidationTest.testRetry(ConnectivityValidationTest.java:190)

   更新： 还不能简单地在checkConnectivity创建的task里面等待新任务结束，那样就违背了checkConnectivity的设计初衷，为了避免长时间占用connectivityExecutor，checkConnectivity采用的是分批处理的策略，所以PR #9211 最终没有修改checkConnectivity方法，而是修改了ConnectivityValidationTest的waitRefresh方法，将等待条件改为invokersToReconnect列表里没有需要进行recover的invoker。