[jira] [Created] (HDFS-12971) DfsClint hang on hedged getFromOneDataNode

["muyannian (JIRA)" <ji...@apache.org>]

muyannian created HDFS-12971:
--------------------------------

             Summary: DfsClint hang on hedged getFromOneDataNode
                 Key: HDFS-12971
                 URL: https://issues.apache.org/jira/browse/HDFS-12971
             Project: Hadoop HDFS
          Issue Type: Bug
          Components: hdfs, hdfs-client
    Affects Versions: 2.7.0, 2.6.0
         Environment: hadoop 2.6.0  
	<property>
	  <name>dfs.client.hedged.read.threadpool.size</name>
	  <value>512</value> 
	</property>
	
	<property>
	  <name>dfs.client.hedged.read.threshold.millis</name>
	  <value>300</value> 
	</property>
            Reporter: muyannian
         Attachments: 1.jpg, 2.jpg, 3.jpg, 4.jpg, DFSInputStream-afterpatch.java, DFSInputStream-beforepatch.java, DFSInputStream.java.patch

when i used hdfs  hedged read , i found DFSInputStream hang on getFirstToComplete method.
the reason is when something has exception on datanode,or namenode , for example FileNotFoundException, that may cause hang up for ever. future has finished but still call get "future = hedgedService.take()" method ,so cause hang up.
in the attatch file is my jstack and java patch.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-help@hadoop.apache.org