-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/60369/#review179499
-----------------------------------------------------------




src/slave/http.cpp
Lines 1330-1334 (patched)
<https://reviews.apache.org/r/60369/#comment254243>

    Whoops, I should have noticed this earlier, but when going to commit this I 
was taking a careful look at the code. This isn't accounting for the pending 
tasks here:
    
    
https://github.com/apache/mesos/blob/0d277bb64fa5a4d0b4f741daedf64095beab4773/src/slave/slave.hpp#L904-L905
    
    Unfortunately the logic is a little tricky. We have to include all of the 
pending task resources. Also, for each executor in the pending map, we have to 
add the ExecutorInfo resources only if it's absent from the 'executors' map. 
You'll want to do a read through of how the data structures are used to confirm 
whether I'm right. :)
    
    My suggestion here would be to have an `allocatedResources()` function 
within the `Framework` struct that contains this logic, and here in the http 
handler we just sum `allocatedResources()` across the frameworks.


- Benjamin Mahler


On June 29, 2017, 3:23 p.m., Andrei Budnik wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/60369/
> -----------------------------------------------------------
> 
> (Updated June 29, 2017, 3:23 p.m.)
> 
> 
> Review request for mesos, Benjamin Mahler and haosdent huang.
> 
> 
> Bugs: MESOS-6441
>     https://issues.apache.org/jira/browse/MESOS-6441
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> The JSON key for this information is "reserved_resources_allocated"
> and "unreserved_resources_allocated".
> 
> 
> Diffs
> -----
> 
>   src/slave/http.cpp cbbc1dc27cc90bac8d48cbbc84266c3d87490a3c 
> 
> 
> Diff: https://reviews.apache.org/r/60369/diff/4/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Andrei Budnik
> 
>

Reply via email to