Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Incomplete
-
0.20.1
-
None
-
None
-
None
Description
Hi,
Sometimes, some of my jobs (It normally always happens in the reducers and on random basis) will not finish and will run forever. I have to manually fail the task so the task will be started and be finished.
The error log on the node is full of entries like:
java.io.IOException: Error Recovery for block blk_-8036012205502614140_21582139 failed because recovery from primary datanode 192.168.0.3:50011 failed 6 times. Pipeline was 192.168.0.3:50011. Aborting...
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.processDatanodeError(DFSClient.java:2582)
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.access$1600(DFSClient.java:2076)
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream$DataStreamer.run(DFSClient.java:2239)
java.io.IOException: Error Recovery for block blk_-8036012205502614140_21582139 failed because recovery from primary datanode 192.168.0.3:50011 failed 6 times. Pipeline was 192.168.0.3:50011. Aborting...
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.processDatanodeError(DFSClient.java:2582)
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.access$1600(DFSClient.java:2076)
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream$DataStreamer.run(DFSClient.java:2239)
java.io.IOException: Error Recovery for block blk_-8036012205502614140_21582139 failed because recovery from primary datanode 192.168.0.3:50011 failed 6 times. Pipeline was 192.168.0.3:50011. Aborting...
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.processDatanodeError(DFSClient.java:2582)
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.access$1600(DFSClient.java:2076)
at org.apache.hadoop.hdfs.DFSClient$DFSOutputStream$DataStreamer.run(DFSClient.java:2239)
The error entries all refer to the same data block.
Unfortunately, the reduce function still seems to be called in the reducer with valid data (although very very slowly), so the task will never been killed and restarted and will take forever to run!
If I kill the task, the job will finish without any problems.
I experienced the same problem under version 0.20.0 as well.
Thanks,
Thibaut