-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/19090/#review37343
-----------------------------------------------------------

Ship it!


Looks good. Just a couple of questions/suggestions.


src/master/drf_sorter.hpp
<https://reviews.apache.org/r/19090/#comment68865>

    # of times since when? Since master/framework was (re)started? Or 
would/could this get recovered on master failover? This gets reset on framework 
restart/failover.
    Would we prefer a rolling window of "# of times client has been chosen for 
allocation in the last X minutes"?



src/master/drf_sorter.cpp
<https://reviews.apache.org/r/19090/#comment68866>

    Do we really want to reset allocations on framework deactivate/reactivate? 
Would we rather save the allocations value and revive it on reactivate?
    Otherwise, a network disconnect or framework restart/failover will allow a 
framework to reset its allocations and move to the top of the list (but only if 
shares are already equal). Not a big deal.


- Adam B


On March 11, 2014, 10:04 p.m., Benjamin Hindman wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/19090/
> -----------------------------------------------------------
> 
> (Updated March 11, 2014, 10:04 p.m.)
> 
> 
> Review request for mesos, Ben Mahler and Vinod Kone.
> 
> 
> Bugs: MESOS-1086
>     https://issues.apache.org/jira/browse/MESOS-1086
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> See bug for more description.
> 
> 
> Diffs
> -----
> 
>   src/master/drf_sorter.hpp f3fab71424fa10f87c87b733cc678fd90579a814 
>   src/master/drf_sorter.cpp bbf8270603e26a1b63e0f96bd5aa9c4ee4eb6d25 
> 
> Diff: https://reviews.apache.org/r/19090/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Benjamin Hindman
> 
>

Reply via email to