mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vinod Kone" <vinodk...@gmail.com>
Subject Re: Review Request 36078: Refactored Call message to include Subscribe message.
Date Wed, 01 Jul 2015 19:22:49 GMT


> On July 1, 2015, 7:06 a.m., Kapil Arya wrote:
> > include/mesos/scheduler/scheduler.proto, line 188
> > <https://reviews.apache.org/r/36078/diff/1/?file=996383#file996383line188>
> >
> >     Should there be a default behavior (say always override the older framework)?
> >     
> >     I also wonder if "override_existing_cheduler" will be a bit clearer than "force"!

I think it is difficult to suggest a default behavior. It depends on the leader election algorithm
and partition semantics that a framework uses. Since it relates to safety I'm a bit reluctant
to prescribe a default.

Regarding the name, I think people grok 'force' easily because it is a well understood term
(esp. in CLIs).


> On July 1, 2015, 7:06 a.m., Kapil Arya wrote:
> > include/mesos/scheduler/scheduler.proto, lines 288-290
> > <https://reviews.apache.org/r/36078/diff/1/?file=996383#file996383line288>
> >
> >     Isn't it true that Call::subscribe().framework_info().id() and Call.framework_id()
won't be set in a given Call message? 
> >     
> >     AFAICT, we do not check whether these two values are set in a given Call message
and if set, whether they are equal or not.

They could both be conceivably set during the Subscribe call by the scheduler. I'll add the
check in scheduler.cpp.


> On July 1, 2015, 7:06 a.m., Kapil Arya wrote:
> > include/mesos/scheduler/scheduler.proto, line 292
> > <https://reviews.apache.org/r/36078/diff/1/?file=996383#file996383line292>
> >
> >     How are we going to allow backwards compatibility. Shouldn't we just mark "framework_info"
as optional and keep setting it for older Masters?

see my comment above about compatibility.


- Vinod


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/36078/#review90011
-----------------------------------------------------------


On July 1, 2015, 5:23 a.m., Vinod Kone wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/36078/
> -----------------------------------------------------------
> 
> (Updated July 1, 2015, 5:23 a.m.)
> 
> 
> Review request for mesos, Adam B, Benjamin Hindman, and Ben Mahler.
> 
> 
> Bugs: MESOs-2551
>     https://issues.apache.org/jira/browse/MESOs-2551
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Subscribe message includes 'FrameworkInfo' and 'force'. Top level protobuf has FrameworkID.
> 
> 
> Diffs
> -----
> 
>   include/mesos/scheduler/scheduler.proto 249ec532b53fe428b7e66be4ced8223e66535b49 
>   src/examples/event_call_framework.cpp 63e42bc83ccc0e4085d7619c478e5b010a49098a 
>   src/master/master.cpp 34ce744f84465ecc9aeecd5fdc3d06047a4b7d92 
>   src/sched/sched.cpp 7563abb85819b0b2bc9afdfd810b33c923c2522e 
>   src/scheduler/scheduler.cpp f360e4d062488986b14e3d48d140996e8ed9e7d6 
>   src/tests/scheduler_tests.cpp cbe6c91a1b4f864ceb11cf062da0ada6c9666f9f 
> 
> Diff: https://reviews.apache.org/r/36078/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Vinod Kone
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message