mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jie Yu" <yujie....@gmail.com>
Subject Re: Review Request 34616: Updated allocator to properly handle oversubscribed resources.
Date Sat, 23 May 2015 00:17:46 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/34616/#review85031
-----------------------------------------------------------

Ship it!


LGTM! Could you please add a NOTE somewhere stating that the default allocator provides DRF
on the total pool of both revokable and non-revokable resources. So it's possible that a framework
allocated with many non-revokable offers will unlikely to get revokable resources. We may
want to experiment with using DRF separately for revokable and non-revokable resources in
the future.

- Jie Yu


On May 22, 2015, 9:41 p.m., Vinod Kone wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/34616/
> -----------------------------------------------------------
> 
> (Updated May 22, 2015, 9:41 p.m.)
> 
> 
> Review request for mesos, Jie Yu and Niklas Nielsen.
> 
> 
> Bugs: MESOS-2734
>     https://issues.apache.org/jira/browse/MESOS-2734
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Just ending up resuing the existing variables (Slave.total and Slave.available) to store
oversubscribed resources. The nice thing is that the changes are minimal. Also, we could potentially
reuse updateSlave() to also update slave's total resources in the future.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.hpp 4b36d42b0c4614493562e57c5eac90c6c38ca087

>   src/tests/hierarchical_allocator_tests.cpp 1a43dc72e739f3c55787716d680faa42a7d0d86f

> 
> Diff: https://reviews.apache.org/r/34616/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Vinod Kone
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message