mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Guangya Liu <gyliu...@gmail.com>
Subject Re: Review Request 43881: Added allocator metric for the number of allocations to a framework.
Date Wed, 24 Feb 2016 12:49:17 GMT


> On 二月 24, 2016, 12:53 a.m., Guangya Liu wrote:
> > src/master/allocator/mesos/hierarchical.cpp, line 1354
> > <https://reviews.apache.org/r/43881/diff/1/?file=1265550#file1265550line1354>
> >
> >     The counter may count twice for some cases, one framework may get resource allocation
in both allocation stage1(Quota) and stage2(wDRF).
> 
> Benjamin Bannier wrote:
>     Good point! The way I counted here we wouldn't get the total number of allocations
over the cluster, but instead the summed number of allocations over slaves (i.e., allocating
on two slaves would have been two allocations instead of one). I fixed the logic for that.
>     
>     I cannot see how we would currently allocate one the same slave twice (once towards
quota, once towards fair share) since we quota also sets a (coarse-grained) max; if such a
scenario existed or will exist it should be captured with the current implementation which
counts allocations over the cluster.

There can be such case that one agent be allocated in stage1(wDRF) and stage2(wDRF), stage1
only allocate non revocable and reserved resources for an agent, stage2 may allocate revocable
resources on the same agent.

I think that your latest patch is good enough because it only count the allocations when allocator
is sending out allocaitons.


- Guangya


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43881/#review120422
-----------------------------------------------------------


On 二月 24, 2016, 10:49 a.m., Benjamin Bannier wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/43881/
> -----------------------------------------------------------
> 
> (Updated 二月 24, 2016, 10:49 a.m.)
> 
> 
> Review request for mesos, Alexander Rukletsov and Ben Mahler.
> 
> 
> Bugs: MESOS-4719
>     https://issues.apache.org/jira/browse/MESOS-4719
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Added allocator metric for the number of allocations to a framework.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.hpp 3043888630b066505410d3b32c5b3f813cc458c1

>   src/master/allocator/mesos/hierarchical.cpp 5ef29f26ec8071f79c2f4f78dbe2bb0a613cc92d

>   src/tests/hierarchical_allocator_tests.cpp 5f771f02db9bd098f3cd36730cd84bf2f5e87a33

> 
> Diff: https://reviews.apache.org/r/43881/diff/
> 
> 
> Testing
> -------
> 
> make check (OS X)
> 
> I confirmed that this does not lead to general performance regressions in the allocator;
this is partially expected since the added code only inserts metrics in the allocator while
the actual work is perform asynchronously. These tests where performed with `HierarchicalAllocator_BENCHMARK_Test.DeclineOffers`
on an optimized build under OS X using clang(trunk) as compiler.
> 
> 
> Thanks,
> 
> Benjamin Bannier
> 
>


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