> On Aug. 31, 2015, 7:49 a.m., Jian Qiu wrote:
> > src/master/master.cpp, line 4904
> > <https://reviews.apache.org/r/37180/diff/6/?file=1052944#file1052944line4904>
> >
> >     If the allocator calls inverse callback followed by an offer callback, 
> > will the two messages be handled in two separate threads of the framework? 
> > I think it will be framework's responsibility to ensure data 
> > synchronization between the two threads?

This is up to how the framework driver is implemented. We have a reference 
implementation, but with the V1 API there will be many implementations.
The current C++ scheduler driver handles 1 message type at a time, so they will 
be synchronized. If the framework decides to do asynchronous calls during those 
callback invocations, then it is up to the framework writer to synchronize them.
Does this make sense?
Since I'm closing this review feel free to reach out to me via e-mail to 
continue this discussion!


- Joris


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/37180/#review97064
-----------------------------------------------------------


On Sept. 13, 2015, 8:33 p.m., Joris Van Remoortere wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/37180/
> -----------------------------------------------------------
> 
> (Updated Sept. 13, 2015, 8:33 p.m.)
> 
> 
> Review request for mesos, Benjamin Hindman, Artem Harutyunyan, and Joseph Wu.
> 
> 
> Bugs: MESOS-1474
>     https://issues.apache.org/jira/browse/MESOS-1474
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   src/master/master.hpp 12cc1ad45de3291ec22d4fe2b7ee11c4d7565c24 
>   src/master/master.cpp c90311fa2152810e7604a0a2dee630bd14929574 
> 
> Diff: https://reviews.apache.org/r/37180/diff/
> 
> 
> Testing
> -------
> 
> The tests break as expected.
> With the scheduler API change there are CHECKs that fail.
> Once we update the API these will be resolved.
> 
> 
> Thanks,
> 
> Joris Van Remoortere
> 
>

Reply via email to