hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Raghu Angadi (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-3164) Use FileChannel.transferTo() when data is read from DataNode.
Date Fri, 04 Apr 2008 01:47:27 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-3164?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12585370#action_12585370
] 

Raghu Angadi commented on HADOOP-3164:
--------------------------------------


bummer... transferTo does not handle non-blocking sockets well. I have a patch that works
fine when the socket channel is in blocking mode. JavaDoc clearly indicates it can handle
non-blocking sockets!. One hack is to interprete IOException and look for "Resource temporarily"
at the beginning of the message.

I get the following with non-blocking sockets :
{noformat}
java.io.IOException: Resource temporarily unavailable
        at sun.nio.ch.FileChannelImpl.transferTo0(Native Method)
        at sun.nio.ch.FileChannelImpl.transferToDirectly(FileChannelImpl.java:418)
        at sun.nio.ch.FileChannelImpl.transferTo(FileChannelImpl.java:519)
        at org.apache.hadoop.dfs.DataNode.transferToFully(DataNode.java:1500)
        at org.apache.hadoop.dfs.DataNode.access$900(DataNode.java:84)
        at org.apache.hadoop.dfs.DataNode$BlockSender.sendChunks(DataNode.java:1807)
        at org.apache.hadoop.dfs.DataNode$BlockSender.sendBlock(DataNode.java:1880)
        at org.apache.hadoop.dfs.DataNode$DataXceiver.readBlock(DataNode.java:1032)
        at org.apache.hadoop.dfs.DataNode$DataXceiver.run(DataNode.java:961)
        at java.lang.Thread.run(Thread.java:619)
{noformat}

Looks like all it needs to do is just to look for EAGAIN from the system call.

> Use FileChannel.transferTo() when data is read from DataNode.
> -------------------------------------------------------------
>
>                 Key: HADOOP-3164
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3164
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: dfs
>            Reporter: Raghu Angadi
>            Assignee: Raghu Angadi
>
> HADOOP-2312 talks about using FileChannel's [{{transferTo()}}|http://java.sun.com/javase/6/docs/api/java/nio/channels/FileChannel.html#transferTo(long,%20long,%20java.nio.channels.WritableByteChannel)]
and [{{transferFrom()}}|http://java.sun.com/javase/6/docs/api/java/nio/channels/FileChannel.html#transferFrom(java.nio.channels.ReadableByteChannel,%20long,%20long)]
in DataNode. 
> At the time DataNode neither used NIO sockets nor wrote large chunks of contiguous block
data to socket. Hadoop 0.17 does both when data is seved to clients (and other datanodes).
I am planning to try using transferTo() in the trunk. This might reduce DataNode's cpu by
another 50% or more.
> Once HADOOP-1702 is committed, we can look into using transferFrom().

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message