> On June 20, 2013, 7:30 p.m., Ben Mahler wrote:
> > hadoop/mesos/src/java/org/apache/hadoop/mapred/MesosScheduler.java, lines 
> > 346-356
> > <https://reviews.apache.org/r/11976/diff/1/?file=308464#file308464line346>
> >
> >     Would it be simpler to think of the minimum in terms of task trackers? 
> > Or is there a reason for cluster specific map/reduce slot minimums?

You don't necessarily have a fixed number of slots per tracker.  It can vary 
with the size of the instance, or the amount of resources available.


- Brenden


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/11976/#review22182
-----------------------------------------------------------


On June 19, 2013, 5:50 p.m., Brenden Matthews wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/11976/
> -----------------------------------------------------------
> 
> (Updated June 19, 2013, 5:50 p.m.)
> 
> 
> Review request for mesos.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Allow specification of mininum slot count.
> 
> To ensure Hadoop jobs begin promptly, we can specify a minimum number of
> 'hot slots' to be available for use.  This addresses the TaskTracker
> spin up delay that exists with Hadoop on Mesos.  This can be a nuisance
> with lower latency applications, such as ad-hoc Hive queries.
> 
> 
> Diffs
> -----
> 
>   hadoop/mesos/src/java/org/apache/hadoop/mapred/MesosScheduler.java 
> c5a63bc7a172c44cc11d6bdd6b2eb4bc3df35cb7 
> 
> Diff: https://reviews.apache.org/r/11976/diff/
> 
> 
> Testing
> -------
> 
> make check && make hadoop
> 
> 
> Thanks,
> 
> Brenden Matthews
> 
>

Reply via email to