> On Jan. 21, 2015, 11:31 p.m., Ben Mahler wrote:
> > src/master/master.hpp, line 429
> > <https://reviews.apache.org/r/29976/diff/3/?file=828919#file828919line429>
> >
> >     Since this is in other places in the master code, might be nice to 
> > follow up on a sweep of the following:
> >     
> >     s/offeredResources/offered/
> >     s/usedResources/used/

Sure, will follow up with a patch.


> On Jan. 21, 2015, 11:31 p.m., Ben Mahler wrote:
> > src/master/master.cpp, line 2766
> > <https://reviews.apache.org/r/29976/diff/3/?file=828920#file828920line2766>
> >
> >     Not yours, mind doing s/offerValidators/validators/ ?
> >     
> >     Seems consistent with the singular 'validator' below as well.
> >     
> >     Feel free to do this as a follow up.

Done.


> On Jan. 21, 2015, 11:31 p.m., Ben Mahler wrote:
> > src/master/master.cpp, lines 2804-2805
> > <https://reviews.apache.org/r/29976/diff/3/?file=828920#file828920line2804>
> >
> >     Could you leverage `drop` here per my other comment? That way, we'll 
> > log consistently when we drop things, regardless of when we drop them. 
> > We'll also get consistent metrics about dropped messages! Thoughts?
> >     
> >     You can add a `drop` overload for Accept:
> >     ```
> >     drop(accept, frameworkInfo, "Framework cannot be found");
> >     ```
> >     
> >     Consider a TODO on this version of `drop` for dealing with recovering 
> > resources from the offers, notice how this is becoming similar to 
> > scheduler.cpp! :)
> >     
> >     Would really like to have the framework info here for consistent 
> > logging.

Added a TODO. THanks for the comment. I like it.


> On Jan. 21, 2015, 11:31 p.m., Ben Mahler wrote:
> > src/master/master.cpp, lines 2894-2896
> > <https://reviews.apache.org/r/29976/diff/3/?file=828920#file828920line2894>
> >
> >     Ditto for `drop` here.
> >     
> >     ```
> >     drop(accept, frameworkInfo, "Framework cannot be found");
> >     ```
> >     
> >     It's not clear if recovering the resources just below here will ever be 
> > necessary, since we don't know the framework! It seems nice for 
> > `drop(Accept, ...)` to consistently handle resource recovery, if applicable.
> >     
> >     We could even consider `drop(Accept, ...)` also dealing with lost task 
> > notifications, through additional overloads (possibly per operation to 
> > cleanup the cruft in the Offer::Operation::LAUNCH case block below), but 
> > one step at a time, a TODO for this latter part would be great :)

Left a TODO.


- Jie


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/29976/#review69066
-----------------------------------------------------------


On Jan. 21, 2015, 10:29 p.m., Jie Yu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/29976/
> -----------------------------------------------------------
> 
> (Updated Jan. 21, 2015, 10:29 p.m.)
> 
> 
> Review request for mesos, Ben Mahler, Michael Park, and Vinod Kone.
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> Moved launchTasks logics to ACCEPT call handler in master.
> 
> This patch moves all logics of launchTasks in master to the ACCEPT call 
> handler in master, and leaves stubs for other offer operations. We still 
> keeps the launchTasks handler for backwards compatibility. launchTasks will 
> simply create an Accept call message, and invoke the ACCEPT call handler.
> 
> 
> Diffs
> -----
> 
>   src/master/master.hpp a8ce4d023ddea36cb83a2dc7b95abd12342f345a 
>   src/master/master.cpp e9dcca3c92c94f3123519855e238bcef47eeece9 
>   src/tests/resource_offers_tests.cpp 
> d098e7016ac0da7f1d629af099bb1b8fa66da839 
> 
> Diff: https://reviews.apache.org/r/29976/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Jie Yu
> 
>

Reply via email to