[jira] [Commented] (HBASE-19708) Avoid NPE when the RPC listener's accept channel is closed

[Chia-Ping Tsai (JIRA)]

    [ 
https://issues.apache.org/jira/browse/HBASE-19708?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16312409#comment-16312409
 ] 

Chia-Ping Tsai commented on HBASE-19708:
----------------------------------------

Does this change make another NPE when stopping RpcServer? 
{code}
@@ -968,6 +971,7 @@ public class RpcServer implements RpcServerInterface, 
ConfigurationObserver {
         } catch (IOException e) {
           LOG.info(getName() + ": exception in closing listener socket. " + e);
         }
+        acceptChannel = null;
       }
       readPool.shutdownNow();
     }
{code}

Listener#run() also try to close the {{Listener}} after loop. 
{code:title=Listener.java}
    public void run() {
      LOG.info(getName() + ": starting");
      while (running) {
         // bababa
      }

      LOG.info(getName() + ": stopping");

      synchronized (this) {
        try {
          acceptChannel.close();  // here
          selector.close();
        } catch (IOException ignored) {
          if (LOG.isTraceEnabled()) LOG.trace("ignored", ignored);
        }

        selector= null;
        acceptChannel= null;

        // clean up all connections
        while (!connectionList.isEmpty()) {
          closeConnection(connectionList.remove(0));
        }
      }
    }
{code}

No wait exist between listener#interrupt and listener#doStop. It cause the NPE 
if the acceptChannel#close is executed after the listener#doStop.
{code:title=RpcServer.java}
  @Override
  public synchronized void stop() {
    LOG.info("Stopping server on " + port);
    running = false;
    if (authTokenSecretMgr != null) {
      authTokenSecretMgr.stop();
      authTokenSecretMgr = null;
    }
    listener.interrupt();
    listener.doStop();
    responder.interrupt();
    scheduler.stop();
    notifyAll();
  }
{code} 


> Avoid NPE when the RPC listener's accept channel is closed
> ----------------------------------------------------------
>
>                 Key: HBASE-19708
>                 URL: https://issues.apache.org/jira/browse/HBASE-19708
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.98.24
>            Reporter: Andrew Purtell
>            Assignee: Andrew Purtell
>            Priority: Minor
>             Fix For: 1.3.2, 1.4.1, 1.5.0
>
>         Attachments: HBASE-19708-branch-1.patch
>
>
> Rare NPE when the listener's accept channel is closed. We serialize access to 
> related state to avoid a previously fixed related NPE and need to do the same 
> for {{acceptChannel}}. Seen in a 0.98 deploy but I think applicable to later 
> code lines. Let me check.
> Exception in thread "MetadataRpcServer.handler=191,queue=0,port=60020" 
> java.lang.NullPointerException
>       at 
> org.apache.hadoop.hbase.ipc.RpcServer$Listener.getAddress(RpcServer.java:858)
>       at 
> org.apache.hadoop.hbase.ipc.RpcServer.getListenerAddress(RpcServer.java:2338)
>       at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:140)
>       at 
> org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:133)
>       at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:108)
>       at java.lang.Thread.run(Thread.java:745)



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)