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 34729: Updated slave to send total amount of oversubscribed resources.
Date Thu, 28 May 2015 00:04:59 GMT


> On May 27, 2015, 11:59 p.m., Niklas Nielsen wrote:
> > src/messages/messages.proto, lines 339-342
> > <https://reviews.apache.org/r/34729/diff/1/?file=973027#file973027line339>
> >
> >     Would it make sense to send them independently, to distinguish between the two
later on?

That was an option we discussed, but dropped it because we felt it was more confusing to send
both because the master/allocator just needs the total. In the future, when we send slave's
(updated) total regular resources, we will also likely send just one value.

Does that make sense?


> On May 27, 2015, 11:59 p.m., Niklas Nielsen wrote:
> > src/slave/slave.cpp, line 4109
> > <https://reviews.apache.org/r/34729/diff/1/?file=973031#file973031line4109>
> >
> >     What does 'pending on the slave' mean here?

tasks could be pending on slave for example, if 'gc->unschedule()' is in progress or executor
hasn't registered etc.


- Vinod


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/34729/#review85466
-----------------------------------------------------------


On May 27, 2015, 11:11 p.m., Vinod Kone wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/34729/
> -----------------------------------------------------------
> 
> (Updated May 27, 2015, 11:11 p.m.)
> 
> 
> Review request for mesos, Jie Yu, Niklas Nielsen, and Brian Wickman.
> 
> 
> Bugs: MESOS-2770
>     https://issues.apache.org/jira/browse/MESOS-2770
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Per the latest design, the slave needs to send the total amount of oversubscribed resources
(both allocated and available).
> 
> Also renamed the message to UpdateSlave to make it more generic (e.g., can be used in
the future to update slave's total resources too).
> 
> 
> Diffs
> -----
> 
>   src/messages/messages.proto 39dac721fbe5b97842dd5d1d68cc135148ae02a2 
>   src/slave/flags.hpp 6ca59dc9fc748ec738259406642ec17c0752590c 
>   src/slave/flags.cpp a8c7c498d674ca832fa052412a373c9ace4b3fc3 
>   src/slave/slave.hpp 0207eafa914e23e4c72b1e5c4fb43aae6c97049c 
>   src/slave/slave.cpp b4d20294330f791e64a597c67b686aed9de84837 
>   src/tests/oversubscription_tests.cpp 75c25b04c1e6a8e0e7e8fd55440743fe1699af88 
> 
> Diff: https://reviews.apache.org/r/34729/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> Will add a test in a subsequent review when a framework is able to launch a revocable
task.
> 
> 
> Thanks,
> 
> Vinod Kone
> 
>


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