Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Hadoop >> mail # user >>


just add more, continue the above thread:

  protected synchronized long getEstimatedTotalMapOutputSize()  {
    if(completedMapsUpdates < threshholdToUse) {
      return 0;
    } else {
      long inputSize = job.getInputLength() + job.desiredMaps();
      //add desiredMaps() so that randomwriter case doesn't blow up
      //the multiplication might lead to overflow, casting it with
      //double prevents it
      long estimate = Math.round(((double)inputSize *
          completedMapsOutputSize * 2.0)/completedMapsInputSize);
      if (LOG.isDebugEnabled()) {
        LOG.debug("estimate total map output will be " + estimate);
      }
      return estimate;
    }
  }
On Sun, Jun 2, 2013 at 12:34 AM, Azuryy Yu <[EMAIL PROTECTED]> wrote:

> This should be fixed in hadoop-1.1.2 stable release.
> if we determine completedMapsInputSize is zero, then job's map tasks MUST
> be zero, so the estimated output size is zero.
> below is the code:
>
>   long getEstimatedMapOutputSize() {
>     long estimate = 0L;
>     if (job.desiredMaps() > 0) {
>       estimate = getEstimatedTotalMapOutputSize()  / job.desiredMaps();
>     }
>     return estimate;
>   }
>
>
>
> On Sat, Jun 1, 2013 at 11:49 PM, Harsh J <[EMAIL PROTECTED]> wrote:
>
>> Does smell like a bug as that number you get is simply Long.MAX_VALUE,
>> or 8 exbibytes.
>>
>> Looking at the sources, this turns out to be a rather funny Java issue
>> (there's a divide by zero happening and [1] suggests Long.MAX_VALUE
>> return in such a case). I've logged a bug report for this at
>> https://issues.apache.org/jira/browse/MAPREDUCE-5288 with a
>> reproducible case.
>>
>> Does this happen consistently for you?
>>
>> [1]
>> http://docs.oracle.com/javase/6/docs/api/java/lang/Math.html#round(double)
>>
>> On Sat, Jun 1, 2013 at 7:27 PM, Lanati, Matteo <[EMAIL PROTECTED]>
>> wrote:
>> > Hi all,
>> >
>> > I stumbled upon this problem as well while trying to run the default
>> wordcount shipped with Hadoop 1.2.0. My testbed is made up of 2 virtual
>> machines: Debian 7, Oracle Java 7, 2 GB RAM, 25 GB hard disk. One node is
>> used as JT+NN, the other as TT+DN. Security is enabled. The input file is
>> about 600 kB and the error is
>> >
>> > 2013-06-01 12:22:51,999 WARN org.apache.hadoop.mapred.JobInProgress: No
>> room for map task. Node 10.156.120.49 has 22854692864 bytes free; but we
>> expect map to take 9223372036854775807
>> >
>> > The logfile is attached, together with the configuration files. The
>> version I'm using is
>> >
>> > Hadoop 1.2.0
>> > Subversion
>> https://svn.apache.org/repos/asf/hadoop/common/branches/branch-1.2 -r
>> 1479473
>> > Compiled by hortonfo on Mon May  6 06:59:37 UTC 2013
>> > From source with checksum 2e0dac51ede113c1f2ca8e7d82fb3405
>> > This command was run using
>> /home/lu95jib/hadoop-exmpl/hadoop-1.2.0/hadoop-core-1.2.0.jar
>> >
>> > If I run the default configuration (i.e. no securty), then the job
>> succeeds.
>> >
>> > Is there something missing in how I set up my nodes? How is it possible
>> that the envisaged value for the needed space is so big?
>> >
>> > Thanks in advance.
>> >
>> > Matteo
>> >
>> >
>> >
>> >>Which version of Hadoop are you using. A quick search shows me a bug
>> >>https://issues.apache.org/jira/browse/HADOOP-5241 that seems to show
>> >>similar symptoms. However, that was fixed a long while ago.
>> >>
>> >>
>> >>On Sat, Mar 23, 2013 at 4:40 PM, Redwane belmaati cherkaoui <
>> >>[EMAIL PROTECTED]> wrote:
>> >>
>> >>> This the content of the jobtracker log file :
>> >>> 2013-03-23 12:06:48,912 INFO org.apache.hadoop.mapred.JobInProgress:
>> Input
>> >>> size for job job_201303231139_0001 = 6950001. Number of splits = 7
>> >>> 2013-03-23 12:06:48,925 INFO org.apache.hadoop.mapred.JobInProgress:
>> >>> tip:task_201303231139_0001_m_000000 has split on
>> >>> node:/default-rack/hadoop0.novalocal
>> >>> 2013-03-23 12:06:48,927 INFO org.apache.hadoop.mapred.JobInProgress:
>> >>> tip:task_201303231139_0001_m_000001 has split on