mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alexander Rukletsov" <ruklet...@gmail.com>
Subject Re: Review Request 42289: Calcuated 'remainingClusterResources' by all activated slaves.
Date Tue, 19 Jan 2016 14:53:33 GMT


> On Jan. 19, 2016, 11:53 a.m., Guangya Liu wrote:
> > src/tests/hierarchical_allocator_tests.cpp, lines 1817-1829
> > <https://reviews.apache.org/r/42289/diff/3/?file=1201483#file1201483line1817>
> >
> >     What about make the test cases cover both `addSlave()` before and after `createFrameworkInfo`?
> 
> Klaus Ma wrote:
>     The case after `createFrameworkInfo` is enough to cover both cases.
> 
> Guangya Liu wrote:
>     Because this test is testing the cases that the resources added after create framework
can be also allocated to framework, I think that it is better to keep `addSlave` cases `before`
and `after` create framework to make sure that the resources can be allocated to framework
for both conditions.

If we add an agent *before* adding any frameworks, we may lay away resources, but no allocations
will be made. Currently we can't really look into the allocator and check whether the resources
have been actually laid away. However, once we have at least one framework, allocations can
be made. Hence I would agree with Klaus that one test case is enough, but we want to make
sure people understand that the order is important and that we actually test two things with
one test.

I'll think about how we can improve comments around the test.


- Alexander


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/42289/#review115143
-----------------------------------------------------------


On Jan. 19, 2016, 2:25 p.m., Klaus Ma wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/42289/
> -----------------------------------------------------------
> 
> (Updated Jan. 19, 2016, 2:25 p.m.)
> 
> 
> Review request for mesos, Alexander Rukletsov, Ben Mahler, Joris Van Remoortere, and
Neil Conway.
> 
> 
> Bugs: MESOS-4102
>     https://issues.apache.org/jira/browse/MESOS-4102
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> __Phenomenon__:
> Quota doesn't allocate resources on slave joining.
> 
> __Root Cause__:
> Event-triggered allocations do not include all available agents. If we
> calculate remaining resources in the cluster using the partial view,
> we may overlook already laid away resources for quota'ed roles and lay
> away more. Hence we may unnecessarily deprive non-quota'ed frameworks
> of resources.
> 
> Refer to AlexR's comments for more detail:
> https://issues.apache.org/jira/browse/MESOS-4102?focusedCommentId=15048495&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15048495
> 
> __Solution/Fix__:
> Calcuated 'remainingClusterResources' by all activated slaves.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.cpp 48acde69b1a2f305b568a7e322a58708063dd30a

>   src/tests/hierarchical_allocator_tests.cpp 9362dd306497ba01e0f387c3862456cdcac6f863

> 
> Diff: https://reviews.apache.org/r/42289/diff/
> 
> 
> Testing
> -------
> 
> make
> make check
> Mac & Ubuntu: ./bin/mesos-tests.sh --gtest_filter="HierarchicalAllocatorTest.*" --verbose
--gtest_repeat=100 --gtest_shuffle
> 
> 
> Thanks,
> 
> Klaus Ma
> 
>


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