> On Oct. 26, 2015, 2 p.m., Klaus Ma wrote:
> > src/master/allocator/mesos/hierarchical.cpp, line 1018
> > <https://reviews.apache.org/r/39401/diff/5/?file=1105060#file1105060line1018>
> >
> >     Can we move this out of the for loop? if the role is satisfied, 
> > framewor sort is not necessary.

And what if it is not but it got satisfied after we allocated resources to 
some—but not all!—frameworks in this role?


> On Oct. 26, 2015, 2 p.m., Klaus Ma wrote:
> > src/master/allocator/mesos/hierarchical.cpp, line 1094
> > <https://reviews.apache.org/r/39401/diff/5/?file=1105060#file1105060line1094>
> >
> >     `allocable(...)` checked whether resource is enough to allocate (cpu > 
> > MIN_CPU && men > MIN_MEN). So if the resources can not allocate here, is 
> > also can not allocate in DRF stage.

But `resources` may differ: during DRF a framework may get some revocable 
resources which may render the whole offer allocatable.


> On Oct. 26, 2015, 2 p.m., Klaus Ma wrote:
> > src/master/allocator/mesos/hierarchical.cpp, line 1101
> > <https://reviews.apache.org/r/39401/diff/5/?file=1105060#file1105060line1101>
> >
> >     My understading of this code is to lay aside resources for unsatisfaied 
> > Quota in next allocation cycle, right?
> >     If so, I think we can move this out to another `foreach slaveIds` loop, 
> > so we did not need to lay aside resources.

Im not sure I got your proposal. We should bookkeep resources that are 
"reserved" or "laid aside" as part of role's quota in order not to offer them 
to anybody else as non-revocable. However, we may want to offer them as 
revocable offers.


- Alexander


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/39401/#review103983
-----------------------------------------------------------


On Oct. 23, 2015, 4:38 p.m., Alexander Rukletsov wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/39401/
> -----------------------------------------------------------
> 
> (Updated Oct. 23, 2015, 4:38 p.m.)
> 
> 
> Review request for mesos, Bernd Mathiske, Joerg Schad, and Joris Van 
> Remoortere.
> 
> 
> Bugs: MESOS-3718
>     https://issues.apache.org/jira/browse/MESOS-3718
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.cpp 
> f4e4a123d3da0442e8b0b0ad14d1ee760752ba36 
> 
> Diff: https://reviews.apache.org/r/39401/diff/
> 
> 
> Testing
> -------
> 
> make check (Mac OS X 10.10.4)
> 
> 
> Thanks,
> 
> Alexander Rukletsov
> 
>

Reply via email to