> On Jan. 13, 2015, 11:29 p.m., Vinod Kone wrote:
> > src/master/master.cpp, line 1337
> > <https://reviews.apache.org/r/29869/diff/1/?file=820049#file820049line1337>
> >
> >     We should consider using 'OneOf' protobuf type to get this check for 
> > free. 
> >     
> >     https://developers.google.com/protocol-buffers/docs/proto#oneof
> 
> Ben Mahler wrote:
>     That's a 2.6.0 feature: 
> https://github.com/google/protobuf/releases/tag/v2.6.0

aha. was wondering why we haven't used it before.


> On Jan. 13, 2015, 11:29 p.m., Vinod Kone wrote:
> > src/master/master.cpp, lines 1380-1381
> > <https://reviews.apache.org/r/29869/diff/1/?file=820049#file820049line1380>
> >
> >     Curious, why you want acceptOffers() call into this method instead of 
> > the current way. I imagine each operation to be implemented as a 
> > function/method, similar to what you did in Master::call().
> 
> Ben Mahler wrote:
>     Yep, that's what we're aiming for, we're thinking of running through the 
> operations and creating a chain of computation.

maybe i'm reading the comment wrong, but isn't the end goal here not to remove 
Master::launchTasks() method but to remove the installation of 
LaunchTasksMessage? in my mind, the fully implemented receive() will look like 
this:

```
void Master::receive(   
    const UPID& from,
    const FrameworkInfo& frameworkInfo,
    const scheduler::Call::Accept& accept) {
    
    for (const Operation operation, accept.operations()) {
       switch (operation.type()) {
         case Offer::Operation::LAUNCH:
           launchTasks(...);
           break;
         case Offer::Operation::RESERVE:
           reserveResources(...);
           break;
         case Offer::Operation::UNRESERVE:
           unreserveResources(...);
           break;
         case Offer::Operation::Create:
           createVolume(...);
           break;
         case Offer::Operation::Destroy:
           destroyVolume(...);
           break;
       }
    }
}
```


- Vinod


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/29869/#review67970
-----------------------------------------------------------


On Jan. 14, 2015, 1:25 a.m., Ben Mahler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/29869/
> -----------------------------------------------------------
> 
> (Updated Jan. 14, 2015, 1:25 a.m.)
> 
> 
> Review request for mesos, Jie Yu, Michael Park, and Vinod Kone.
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> This is an initial stub that only handles accepting offers.
> 
> The plan (per the TODO) is to have a single path for launching tasks through 
> `acceptOffers`, as opposed to `launchTasks`.
> 
> 
> Diffs
> -----
> 
>   src/master/master.hpp 26116aff1e965501c8d94ea0b5bd1be37f944887 
>   src/master/master.cpp 63ca19ab9618feccd93a2335f9287122a4665c5e 
> 
> Diff: https://reviews.apache.org/r/29869/diff/
> 
> 
> Testing
> -------
> 
> The code path currently cannot be executed. Tests will be added once the 
> scheduler driver supports accepting offers.
> 
> make check
> 
> 
> Thanks,
> 
> Ben Mahler
> 
>

Reply via email to