Nartan, If you're using BytesWritable, I've heard that it doesn't return
only valid bytes, it actually returns more than that.
Here is this issue discussed:
http://www.nabble.com/can%27t-read-the-SequenceFile-correctly-td21866960.html

Cheers,
Rasit



2009/2/18 Nathan Marz <nat...@rapleaf.com>

> Hello,
>
> I'm seeing very odd numbers from the HDFS job tracker page. I have a job
> that operates over approximately 200 GB of data (209715200047 bytes to be
> exact), and HDFS bytes read is 2,103,170,802,501 (2 TB).
>
> The "Map input bytes" is set to "209,714,811,510", which is a correct
> number.
>
> The job only took 10 minutes to run, so there's no way that that much data
> was actually read. Anyone have any idea of what's going on here?
>
> Thanks,
> Nathan Marz
>



-- 
M. Raşit ÖZDAŞ

Reply via email to