Yes. Hdfs and Mapred related dirs are set outside of /tmp.

On Tue, Dec 27, 2011 at 6:48 PM, alo alt <[email protected]> wrote:

> Hi,
>
> did you set the hdfs-related dirs outside of /tmp? Most *ux systems
> clean them up on reboot.
>
> - Alex
>
> On Tue, Dec 27, 2011 at 2:09 PM, Rajat Goel <[email protected]> wrote:
> > Hi,
> >
> > I have a 7-node setup (1 - Namenode/JobTracker, 6 -
> Datanodes/TaskTrackers)
> > running Hadoop version 0.20.203.
> >
> > I performed the following test:
> > Initially cluster is running smoothly. Just before launching a MapReduce
> > job (about one or two minutes before), I shutdown one of the data nodes
> > (rebooted the machine). Then my MapReduce job starts but immediately
> fails
> > with following messages on stderr:
> >
> > WARNING: org.apache.hadoop.metrics.jvm.EventCounter is deprecated. Please
> > use org.apache.hadoop.log.metrics.EventCounter in all the
> log4j.properties
> > files.
> > WARNING: org.apache.hadoop.metrics.jvm.EventCounter is deprecated. Please
> > use org.apache.hadoop.log.metrics.EventCounter in all the
> log4j.properties
> > files.
> > WARNING: org.apache.hadoop.metrics.jvm.EventCounter is deprecated. Please
> > use org.apache.hadoop.log.metrics.EventCounter in all the
> log4j.properties
> > files.
> > WARNING: org.apache.hadoop.metrics.jvm.EventCounter is deprecated. Please
> > use org.apache.hadoop.log.metrics.EventCounter in all the
> log4j.properties
> > files.
> > NOTICE: Configuration: /device.map    /region.map    /url.map
> > /data/output/2011/12/26/08
> >  PS:192.168.100.206:11111    3600    true    Notice
> > 11/12/26 09:10:26 WARN mapred.JobClient: Use GenericOptionsParser for
> > parsing the arguments. Applications should implement Tool for the same.
> > 11/12/26 09:10:26 INFO input.FileInputFormat: Total input paths to
> process
> > : 24
> > 11/12/26 09:10:37 INFO hdfs.DFSClient: Exception in
> createBlockOutputStream
> > java.io.IOException: Bad connect ack with firstBadLink as
> > 192.168.100.5:50010
> > 11/12/26 09:10:37 INFO hdfs.DFSClient: Abandoning block
> > blk_-6309642664478517067_35619
> > 11/12/26 09:10:37 INFO hdfs.DFSClient: Waiting to find target node:
> > 192.168.100.7:50010
> > 11/12/26 09:10:44 INFO hdfs.DFSClient: Exception in
> createBlockOutputStream
> > java.net.NoRouteToHostException: No route to host
> > 11/12/26 09:10:44 INFO hdfs.DFSClient: Abandoning block
> > blk_4129088682008611797_35619
> > 11/12/26 09:10:53 INFO hdfs.DFSClient: Exception in
> createBlockOutputStream
> > java.io.IOException: Bad connect ack with firstBadLink as
> > 192.168.100.5:50010
> > 11/12/26 09:10:53 INFO hdfs.DFSClient: Abandoning block
> > blk_3596375242483863157_35619
> > 11/12/26 09:11:01 INFO hdfs.DFSClient: Exception in
> createBlockOutputStream
> > java.io.IOException: Bad connect ack with firstBadLink as
> > 192.168.100.5:50010
> > 11/12/26 09:11:01 INFO hdfs.DFSClient: Abandoning block
> > blk_724369205729364853_35619
> > 11/12/26 09:11:07 WARN hdfs.DFSClient: DataStreamer Exception:
> > java.io.IOException: Unable to create new block.
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.nextBlockOutputStream(DFSClient.java:3002)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.access$2000(DFSClient.java:2255)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream$DataStreamer.run(DFSClient.java:2446)
> >
> > 11/12/26 09:11:07 WARN hdfs.DFSClient: Error Recovery for block
> > blk_724369205729364853_35619 bad datanode[1] nodes == null
> > 11/12/26 09:11:07 WARN hdfs.DFSClient: Could not get block locations.
> > Source file
> >
> "/data/hadoop-admin/mapred/staging/admin/.staging/job_201112200923_0292/job.split"
> > - Aborting...
> > 11/12/26 09:11:07 INFO mapred.JobClient: Cleaning up the staging area
> >
> hdfs://machine-100-205:9000/data/hadoop-admin/mapred/staging/admin/.staging/job_201112200923_0292
> > Exception in thread "main" java.io.IOException: Bad connect ack with
> > firstBadLink as 192.168.100.5:50010
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.createBlockOutputStream(DFSClient.java:3068)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.nextBlockOutputStream(DFSClient.java:2983)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.access$2000(DFSClient.java:2255)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream$DataStreamer.run(DFSClient.java:2446)
> > 11/12/26 09:11:07 ERROR hdfs.DFSClient: Exception closing file
> >
> /data/hadoop-admin/mapred/staging/admin/.staging/job_201112200923_0292/job.split
> > : java.io.IOException: Bad connect ack with firstBadLink as
> > 192.168.100.5:50010
> > java.io.IOException: Bad connect ack with firstBadLink as
> > 192.168.100.5:50010
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.createBlockOutputStream(DFSClient.java:3068)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.nextBlockOutputStream(DFSClient.java:2983)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.access$2000(DFSClient.java:2255)
> >    at
> >
> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream$DataStreamer.run(DFSClient.java:2446)
> >
> >
> > - In the above logs, 192.168.100.5 is the machine I rebooted.
> > - JobTracker's log file doesn't have any logs in the above time period.
> > - NameNode's log file doesn't have any exceptions or any messages related
> > to the above error logs.
> > - All nodes can access each other via IP or hostnames.
> > - ulimit values for files is set to 1024 but I don't see many connections
> > in CLOSE_WAIT state (Googled a bit and some ppl suggest that this value
> > could be a culprit in some cases)
> > - My Hadoop configuration files have settings for no. of mappers (8),
> > reducers (4), io.sort.mb (512 mb). Most of the other parameters have been
> > configured to their default values.
> >
> > Can someone please provide any pointers to solution of this problem?
> >
> > Thanks,
> > Rajat
>
>
>
> --
> Alexander Lorenz
> http://mapredit.blogspot.com
>
> P Think of the environment: please don't print this email unless you
> really need to.
>

Reply via email to