mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Klaus Ma" <klaus1982...@gmail.com>
Subject Re: Review Request 39056: Fix for Mesos master crash due to check failure.
Date Tue, 24 Nov 2015 15:36:04 GMT


> On Oct. 7, 2015, 8:42 a.m., Klaus Ma wrote:
> > src/common/resources.cpp, line 879
> > <https://reviews.apache.org/r/39056/diff/3/?file=1092072#file1092072line879>
> >
> >     This fix is ok for this ticket; but how to handle other part about cpu()? Here's
some question from me:
> >     1. if `epsilon` is 0.01 here, does it mean the min cpu is 0.01?
> >     2. is this the only code that need `epsilon`? It seems not
> >     
> >     @jieyu/@mcypark, show we start a EPIC to include all precision related ticket?
so, we can 
> >     1. unify the min cpu/men/disk to user
> >     2. unify the operator/code within allocator
> >     3. unify the precision between backend and UI
> >     4. clarify the requirement to cpu/mem, e.g. whether accept empty cpu/mem
> >     
> >     Any comments?
> 
> Bernd Mathiske wrote:
>     I suggest (based on input from @benjaminhindman) we don't use a specific constant
here, but try using CHECK_DOUBLE_EQ, first. Only if this does not work, we should switch to
CHECK_NEAR, which takes epsilon as an argument. (In that case, we'd have to elevate the constant
to a more prominent place and describe it in the docs as well.) With either of this, we'll
have a viable short-term fix. The long-term solution will be pursued by MESOS-3997.

Agree that do not use constant here. For the CHECK_NEAR/CHECK_DOUBLE_EQ, suggest to introduce
`almostEqual()` in gtest; because we also need to check whether two Scalar equal.


- Klaus


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/39056/#review101727
-----------------------------------------------------------


On Oct. 7, 2015, 6:07 a.m., Mandeep Chadha wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/39056/
> -----------------------------------------------------------
> 
> (Updated Oct. 7, 2015, 6:07 a.m.)
> 
> 
> Review request for mesos and Neil Conway.
> 
> 
> Bugs: MESOS-3552
>     https://issues.apache.org/jira/browse/MESOS-3552
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Check failed due to double comparison : MESOS-3552.
> 
> 
> Diffs
> -----
> 
>   src/common/resources.cpp 601388c35a1bff37c58e753d1870d53b8d0af2d1 
>   src/tests/reservation_tests.cpp 6b7c43c8b5c64618249dbee926383242320c111e 
>   src/v1/resources.cpp dc868903472f8f3a1ddc56092e3f8f81d953ce39 
> 
> Diff: https://reviews.apache.org/r/39056/diff/
> 
> 
> Testing
> -------
> 
> Added unit test.
> make check successful.
> 
> 
> Thanks,
> 
> Mandeep Chadha
> 
>


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