Yes, I use the new wire protocol.

I am currently working with our legal department to make the code open source.

-h



On Wed, Jul 9, 2014 at 4:04 PM, Adam Bordelon <[email protected]> wrote:
>> "a pure-Java implementation of the framework code"
> Sounds great. Are you using the new wire protocol to avoid the jni/libmesos?
> Would love to see your code.
> -A-
>
>
> On Wed, Jul 9, 2014 at 3:51 PM, Henning Schmiedehausen <
> [email protected]> wrote:
>
>> I am actually working on a pure-Java implementation of the framework code
>> and I was confused that this method is nowhere called.
>>
>> Thanks,
>>     Henning
>>
>>
>> On Wed, Jul 9, 2014 at 3:08 PM, Adam Bordelon <[email protected]> wrote:
>>
>> > ExecutorLost is not called yet. See
>> > https://issues.apache.org/jira/browse/MESOS-313
>> > MESOS-313 - report executor deaths to framework schedulers
>> > Should be a simple patch, but apparently low enough demand that nobody's
>> > done it yet. Care to try?
>> >
>> >
>> > On Wed, Jul 9, 2014 at 2:47 PM, Henning Schmiedehausen <
>> > [email protected]> wrote:
>> >
>> > > Hi,
>> > >
>> > > looking through the Mesos source code, especially sched/sched.cpp,
>> there
>> > > are receivers installed for the following message types, which in turn
>> > > drive calls to the various scheduler methods:
>> > >
>> > > * FrameworkRegisteredMessage drives scheduler-->registered()
>> > > * FrameworkReregisteredMessage drives scheduler->reregistered()
>> > > * ResourceOffersMessage drives scheduler->resourceOffers()
>> > > * RescindResourceOfferMessage drives scheduler->offerRescinded()
>> > > * StatusUpdateMessage drives scheduler->statusUpdate()
>> > > * LostSlaveMessage drives scheduler->slaveLost()
>> > > * ExecutorToFrameworkMessage drives scheduler->frameworkMessage()
>> > > * FrameworkErrorMessage drives scheduler->error()
>> > >
>> > > However, I am unable to find the place which calls the
>> > > scheduler->executorLost() method. What am I missing?
>> > >
>> > > Thanks for helping me out here,
>> > >     Henning
>> > >
>> >
>>

Reply via email to