-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/55355/#review160992
-----------------------------------------------------------



I'd run this change my @jie.


src/slave/containerizer/mesos/io/switchboard.cpp (line 993)
<https://reviews.apache.org/r/55355/#comment232237>

    Can you move this declaration just below the `inputConnected` declaration?
    
    It seems to group better there for me.



src/slave/containerizer/mesos/io/switchboard.cpp (line 1265)
<https://reviews.apache.org/r/55355/#comment232239>

    It's probably worth mentioning here why we grab a reference to the accept 
(i.e. so that we can discard it in `finalize()` so that they FD associated with 
the socket doesn't get leaked).


- Kevin Klues


On Jan. 9, 2017, 8:35 p.m., Greg Mann wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/55355/
> -----------------------------------------------------------
> 
> (Updated Jan. 9, 2017, 8:35 p.m.)
> 
> 
> Review request for mesos, Benjamin Bannier, Benjamin Hindman, Benjamin 
> Mahler, Joseph Wu, and Kevin Klues.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Previously, the IO switchboard could leak file descriptors
> because it held a reference to its server socket within the
> socket's accept loop. This patch explicitly discards the
> future containing this reference to eliminate the leak.
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/mesos/io/switchboard.cpp 
> 67c51678e87389ab08a81fe4832cffd3d83bdae9 
> 
> Diff: https://reviews.apache.org/r/55355/diff/
> 
> 
> Testing
> -------
> 
> Run:
> `bin/mesos-tests.sh --gtest_repeat=-1 
> --gtest_filter="IOSwitchboardServerTest.AttachInput"`
> 
> and then simultaneously execute `sudo lsof -p PID_of_test_process | wc -l` 
> repeatedly to ensure that the number of open FDs is not increasing.
> 
> 
> Thanks,
> 
> Greg Mann
> 
>

Reply via email to