mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Benjamin Bannier <benjamin.bann...@mesosphere.io>
Subject Re: Review Request 43881: Added allocator metric for the number of allocations to a framework.
Date Fri, 04 Mar 2016 16:02:10 GMT


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/master/allocator/mesos/hierarchical.cpp, lines 275-280
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279203#file1279203line275>
> >
> >     I would pull this into a helper in `Metrics` similar to `createGaugesForResource()`.
This way we keep the allocator code minimal.

Done.


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/master/allocator/mesos/hierarchical.cpp, lines 344-347
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279203#file1279203line344>
> >
> >     Maybe here a helper as well.

Done.


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/master/allocator/mesos/hierarchical.cpp, lines 1498-1499
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279203#file1279203line1498>
> >
> >     I think this should be around `allocated()` calls for sorters. When `allocated()`
is called, the sorter's internal allocation counter is incremented. I'd say we should keep
the counters in sync.
> >     
> >     Alternatively, we can expose the allocations counter from sorters and query
it directly.

For now I put these in the locations where a `Sorter`'s `allocated` is called.


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/master/allocator/mesos/metrics.cpp, lines 59-60
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279205#file1279205line59>
> >
> >     If you say `using process::metrics::Counter` at the top, you can fit this into
one line : ).

Done, also for other `Metric`s in introduced here in this series.


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/tests/hierarchical_allocator_tests.cpp, line 2464
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279206#file1279206line2464>
> >
> >     Backtick `framework1` please.

Done.


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/tests/hierarchical_allocator_tests.cpp, lines 2464-2472
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279206#file1279206line2464>
> >
> >     Do you want to set quota as part of your test? Because otherwise we can simplify
the test a bit by not setting it and having `agent2` allocated to `framework2`.

It does not hurt introducing it here, and we will make more explicit use of this information
at a later point in this series.


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/tests/hierarchical_allocator_tests.cpp, line 2487
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279206#file1279206line2487>
> >
> >     Feel free to kill this line

It's dead, Alex :D


> On March 4, 2016, 2:05 p.m., Alexander Rukletsov wrote:
> > src/tests/hierarchical_allocator_tests.cpp, lines 2492-2495
> > <https://reviews.apache.org/r/43881/diff/15/?file=1279206#file1279206line2492>
> >
> >     How about 
> >     ```
> >       Clock::advance(flags.allocation_interval);
> >       Clock::settle();
> >       ++allocations;
> >     ```
> >     to keep clock manipulation together?

So be it.


- Benjamin


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43881/#review122053
-----------------------------------------------------------


On March 4, 2016, 5:01 p.m., Benjamin Bannier wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/43881/
> -----------------------------------------------------------
> 
> (Updated March 4, 2016, 5:01 p.m.)
> 
> 
> Review request for mesos, Alexander Rukletsov and Ben Mahler.
> 
> 
> Bugs: MESOS-4719
>     https://issues.apache.org/jira/browse/MESOS-4719
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   docs/monitoring.md 827f7073204fcf8575ca980a5571c8be4f5e4110 
>   src/master/allocator/mesos/hierarchical.cpp 70291075c00a9a557529c2562dedcfc6c6c3ec32

>   src/master/allocator/mesos/metrics.hpp PRE-CREATION 
>   src/master/allocator/mesos/metrics.cpp PRE-CREATION 
>   src/tests/hierarchical_allocator_tests.cpp 3e4ad31925e1b815a74d67fa3962d23fa5bc89d1

> 
> 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