> On Oct. 21, 2014, 1:07 a.m., Ben Mahler wrote:
> > src/sched/sched.cpp, lines 1018-1021
> > <https://reviews.apache.org/r/26951/diff/1/?file=726110#file726110line1018>
> >
> >     What's the point of exposing this one, there are no calls to route()..?
> 
> Dominic Hamon wrote:
>     For consistency with the Master event queue metrics.
> 
> Dominic Hamon wrote:
>     to think of it another way: if these metrics were exposed at the Process 
> level (where they should be, if we can solve the dependency issue) then the 
> Driver would have the http request metric. Does that help sway you?

When I've been thinking of them at the Process level, I had still been thinking 
of them as being opt-in. Otherwise, we might have some performance issues?

There would be a ton of anonymous processes (e.g. internal libprocess 
Processes, like HttpProxy, RouteProcess, etc.), or unnecessary processes (like 
tens of thousands of SlaveObserver processes) causing metrics to be really 
expensive.


- Ben


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/26951/#review57507
-----------------------------------------------------------


On Oct. 20, 2014, 7:07 p.m., Dominic Hamon wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/26951/
> -----------------------------------------------------------
> 
> (Updated Oct. 20, 2014, 7:07 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-1943
>     https://issues.apache.org/jira/browse/MESOS-1943
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> add dispatch and http request event counters to scheduler driver
> 
> 
> Diffs
> -----
> 
>   src/sched/sched.cpp a37ed3d2e11035650b9bf0440fb87f66669129d8 
>   src/tests/scheduler_tests.cpp d6a7867452d19cf1e65b780d0801cb4a0cd98e18 
> 
> Diff: https://reviews.apache.org/r/26951/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Dominic Hamon
> 
>

Reply via email to