In addition:

" In the course of pig execution, I see that
many jobs (map/reduce pair) are created. Want to understand what each of
these jobs solve."

For each MR job, there is a property in its xml config called
"pig.alias" that tells you which of the relations in your pig script
are involved in this particular job.

D

On Sun, Jul 3, 2011 at 9:11 PM, Daniel Dai <da...@hortonworks.com> wrote:
> On Sun, Jul 3, 2011 at 10:00 PM, Prabhu Dhakshina Murthy <
> prabh...@yahoo-inc.com> wrote:
>
>> I have few questions on running the pig script/ map-reduce jobs.
>>
>> 1. I know that pig creates *logical, physical and then execution plans*
>> before it really starts executing the map/reduce job; I am able to look
>> at the logical/physical plans using the command *explain <alias_name>*;
>> But how do I view the execution plan (which I suppose list the different
>> map/reduce tasks planned)? In the course of pig execution, I see that
>> many jobs (map/reduce pair) are created. Want to understand what each of
>> these jobs solve.
>>
>>  "explain alias" will show logical plan, physical plan and MR plan. Check
> carefully.
>
> 2. Is there any definitive guide which I can use to understand the plans
>> created because what is spat is difficult to understand.
>>
> Check *
> http://ofps.oreilly.com/titles/9781449302641/developing_and_testing.html#dev_tools
> *
>
>
>
> 3. I am able to change the number of map jobs by changing the number of
>> input file blocks. Do I have control over the number of reduce jobs as
>> well? How do I set the number of reducers?
>>
> Check
> http://pig.apache.org/docs/r0.8.1/cookbook.html#Use+the+Parallel+Features
>
>
>> 4. What is the default heap memory size in mapper/reducer nodes? Which
>> job parameters reflect these? Will I be able to change the heap memory
>> by -Xmx 1024m option? My jobs used to fail when I set the heap memory in
>> this way - May be there are some restrictions on what values can be
>> supplied?
>>
> It is controlled "mapred.*child*.java.opts"
>
> Daniel
>
>
>>
>> Thanks much!
>>
>

Reply via email to