hadoop-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From mg <userformailingli...@gmail.com>
Subject Re: Relevance of mapreduce.* configuration properties for MR V2
Date Thu, 16 Aug 2012 17:37:08 GMT
Hi,

thanks.

But there are many more mapreduce.* properties. ...

Does anyone have more information on those with respect to whether they 
are relevant to YARN, and if so, what they exactly mean in the YARN context?

I am trying to optimize settings in order to use as much RAM and CPU as 
possible. The cluster has 8 node managers each with 96GB RAM and 16 CPUs 
(cpuinfo: Intel(R) Xeon(R) CPU E5620 @ 2.40GHz)

Cheers,
Martin

On 16.08.2012 18:40, anil gupta wrote:
> Hi,
>
> AFAIK, these properties are being ignored by YARN:
> - mapreduce.tasktracker.map.tasks.maximum,
> - mapreduce.tasktracker.reduce.tasks.maximum
>
> Thanks,
> Anil Gupta
>
>
> On Thu, Aug 16, 2012 at 9:28 AM, mg <userformailinglists@gmail.com
> <mailto:userformailinglists@gmail.com>> wrote:
>
>     Hi,
>
>     I am currently trying to tune a CDH 4.0.1 (i~ hadoop 2.0.0-alpha)
>     cluster running HDFS, YARN, and HBase managed by Cloudera Manager
>     4.0.3 (Free Edition).
>
>
>     In CM, there are a number of options for setting mapreduce.*
>     configuration properties on the YARN client page.
>
>     Some of the properties and their explanations in the GUI still refer
>     to JobTracker and TaskTracker, e.g.,
>
>     - mapreduce.jobtracker.handler.__count,
>     - mapreduce.tasktracker.map.__tasks.maximum,
>     - mapreduce.tasktracker.reduce.__tasks.maximum
>
>     I wonder whether these and a number of other mapreduce.* (e.g.,
>     mapreduce.job.reduces) properties are observed by the MR2
>     ApplicationMaster (and how), or not.
>
>
>     Can anyone clarify or point to respective documentation?
>
>     Thanks,
>     Martin
>
>
>
>
> --
> Thanks & Regards,
> Anil Gupta

Mime
View raw message