The difference is that same client behaviour results in different exception in 0.16 and 0.17 because of change to use NIO sockets. The current code ignores "SocketException". But with nio sockets, we just get an IOException. I will file a jira to avoid these error messages.

For now these can be ignored.

Raghu.

Brian Karlak wrote:

2008-06-30 19:27:45,760 ERROR org.apache.hadoop.dfs.DataNode: 192.168.23.1:500
10:DataXceiver: java.io.IOException: Connection reset by peer


Hello All --

We also see this behavior. The Hadoop infrastructure appears to handle these exceptions, in so much as the jobs still complete normally, but it is disconcerting to see so many exceptions popping up in the logs.

This behavior appears to have started as soon as we upgraded to 0.17.0. It is still occurring in yesterday's 0.17.1 release. I have not been able to reproduce it in the 0.16.4 or 0.16.3 releases.

I'm a bit of a noob, but I wonder it it is possibly related to HADOOP-2346, the introduction of timeouts on socket writes? Are there any parameters to alter the timeout behavior? Or is the timeout hardcoded?

We are also investigating HADOOP-3051 as a possible factor, considering that the base exception is being raised in the sun.nio.ch package.

This issue is consistent and reproducible in both of our clusters. it appears to occur with high I/O load jobs. For instance, it occurs on both our current production cluster as well as our the new 3-node cluster whenever we run the "sort" test in the example jobs. It does NOT occur when running the "pi" test.

Any clues or leads would be most appreciated.

Thanks,
Brian

On Jun 30, 2008, at 4:30 AM, Rong-en Fan wrote:

Hi,

I'm using Hadoop 0.17.1 with HBase trunk, and notice lots of exception
in hadoop's log (it's a 3-node hdfs):

2008-06-30 19:27:45,760 ERROR org.apache.hadoop.dfs.DataNode: 192.168.23.1:500
10:DataXceiver: java.io.IOException: Connection reset by peer
       at sun.nio.ch.FileDispatcher.write0(Native Method)
       at sun.nio.ch.SocketDispatcher.write(SocketDispatcher.java:29)
       at sun.nio.ch.IOUtil.writeFromNativeBuffer(IOUtil.java:104)
       at sun.nio.ch.IOUtil.write(IOUtil.java:75)
       at sun.nio.ch.SocketChannelImpl.write(SocketChannelImpl.java:334)
at org.apache.hadoop.net.SocketOutputStream$Writer.performIO(SocketOutputStream.java:53) at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:140) at org.apache.hadoop.net.SocketOutputStream.write(SocketOutputStream.java:144) at org.apache.hadoop.net.SocketOutputStream.write(SocketOutputStream.java:105) at java.io.BufferedOutputStream.write(BufferedOutputStream.java:105)
       at java.io.DataOutputStream.write(DataOutputStream.java:90)
at org.apache.hadoop.dfs.DataNode$BlockSender.sendChunks(DataNode.java:1774) at org.apache.hadoop.dfs.DataNode$BlockSender.sendBlock(DataNode.java:1813) at org.apache.hadoop.dfs.DataNode$DataXceiver.readBlock(DataNode.java:1039) at org.apache.hadoop.dfs.DataNode$DataXceiver.run(DataNode.java:968)
       at java.lang.Thread.run(Thread.java:619)

It seems to me that the datanode can not handle the incoming traffic.
If so, what parameters in hadoop sire and/or in os (I'm using rhel 4) that
I can play with?

Thanks,
Rong-En Fan



Reply via email to