hi Steve,
What's your HDFS release version? From the error log and HDFS0.21
code, I guess that the file does not have any replicas. You may focus on
the missing replica of this file.
Pay attention to the NameNode log with that block id and track the
replica distribution. Or check the NameNode 50070 web interface to find out
concrete DataNodes. It may caused by the failure of DataNode. The final
root cause need your tracking process. Good luck
-Regards
Denny Ye
On Wed, Nov 9, 2011 at 2:00 PM, Steve Lewis <lordjoe2...@gmail.com> wrote:
> Just recently my hadoop jobs started failing with Could not obtain
> block. I recently restarted the cluster but this error has killed 3 jobs
> I need some hints on how tp diagnose and fix the problem
>
> d by: java.io.IOException: Could not obtain block:
> blk_8697778223665513111_1917303
> file=/user/howdah/JXTandem/data/VentnerEcoli/tandem256000.xml
> at
> org.apache.hadoop.hdfs.DFSClient$DFSInputStream.chooseDataNode(DFSClient.java:1797)
> at
> org.apache.hadoop.hdfs.DFSClient$DFSInputStream.blockSeekTo(DFSClient.java:1623)
> at
> org.apache.hadoop.hdfs.DFSClient$DFSInputStream.read(DFSClient.java:1752)
> at
> org.apache.hadoop.hdfs.DFSClient$DFSInputStream.read(DFSClient.java:1680)
>
> --
> Steven M. Lewis PhD
> 4221 105th Ave NE
> Kirkland, WA 98033
> 206-384-1340 (cell)
> Skype lordjoe_com
>
>
>
