hadoop-mapreduce-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Shipper, Jay [USA]" <Shipper_...@bah.com>
Subject HDFS Block Bad Response Error
Date Mon, 16 Mar 2015 19:37:28 GMT
On a Hadoop 2.4.0 cluster, I have a job running that's encountering the following warnings
in one of its map tasks (IPs changed, but otherwise, this is verbatim):

---
2015-03-16 06:59:37,994 WARN [ResponseProcessor for block BP-437460642-10.0.0.1-1391018641114:blk_1084609656_11045296]
org.apache.hadoop.hdfs.DFSClient: DFSOutputStream ResponseProcessor exception  for block BP-437460642-10.0.0.1-1391018641114:blk_1084609656_11045296
java.io.EOFException: Premature EOF: no length prefix available
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:1990)
at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.readFields(PipelineAck.java:176)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:796)
2015-03-16 06:59:37,994 WARN [ResponseProcessor for block BP-437460642-10.0.0.1-1391018641114:blk_1084609655_11045295]
org.apache.hadoop.hdfs.DFSClient: DFSOutputStream ResponseProcessor exception  for block BP-437460642-10.0.0.1-1391018641114:blk_1084609655_11045295
java.io.IOException: Bad response ERROR for block BP-437460642-10.0.0.1-1391018641114:blk_1084609655_11045295
from datanode 10.0.0.1:1019
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:819)
---

This job is launched from Hive 0.13.0, and it's consistently happening on the same split,
which is on a sequence file.  After logging a few errors like the above, the map task seems
to make no progress and eventually times out (with a mapreduce.task.timeout value greater
than 5 hours).

Any pointers on how to begin troubleshooting and resolving this issue?  In searching around,
it was suggested that this is indicative of a "network issue", but as it happens on the same
split consistently, that explanation seems unlikely.

Mime
View raw message