mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gilbert Song <songzihao1...@gmail.com>
Subject Re: Review Request 52235: Supported mesos containerizer recover to be nested aware.
Date Mon, 26 Sep 2016 17:14:57 GMT


> On Sept. 25, 2016, 12:27 p.m., Jie Yu wrote:
> > src/slave/containerizer/mesos/containerizer.cpp, line 705
> > <https://reviews.apache.org/r/52235/diff/1/?file=1509681#file1509681line705>
> >
> >     Why the extra `!containerId.has_parent` check here? Will the first check implie
the second check?

Just put a more accurate check on purpose


> On Sept. 25, 2016, 12:27 p.m., Jie Yu wrote:
> > src/slave/containerizer/mesos/containerizer.cpp, lines 800-812
> > <https://reviews.apache.org/r/52235/diff/1/?file=1509681#file1509681line800>
> >
> >     Why having a separate look for extras? Can you just inline into the first loop?
Let's keep creating of `Container` and insert into its parent close together.

I did it on purpose, since it may be possible that the extra list from the `launcher->recover()`
container parent-child relationship. However, that is a hashset. We cannot guarantee when
we are adding the `containerId` to its parent children list, the parentContainerId exists
in `containers_` already.


> On Sept. 25, 2016, 12:27 p.m., Jie Yu wrote:
> > src/slave/containerizer/mesos/containerizer.cpp, lines 743-748
> > <https://reviews.apache.org/r/52235/diff/1/?file=1509681#file1509681line743>
> >
> >     We still need to call 'destroy' for a container that does not have pid to make
sure it is properly cleaned up.
> >     
> >     ```
> >     container->status = pid.isSome()
> >       ? reap(containerId, pid.get())
> >       : Future<Option<int>>(None());
> >       
> >     container->status->onAny(defer(self(), &Self::reaped, containerId));
> >     ```

I handled the two cases as orphans:
1. no pid containers.
2. orphans from in the `extra` list from launcher->recover().

they are cleaned up at the end of contaienrizer::_recover(). why do we call `reaped()` and
destroy twice?


- Gilbert


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/52235/#review150344
-----------------------------------------------------------


On Sept. 24, 2016, 11:50 a.m., Gilbert Song wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/52235/
> -----------------------------------------------------------
> 
> (Updated Sept. 24, 2016, 11:50 a.m.)
> 
> 
> Review request for mesos, Benjamin Hindman, Artem Harutyunyan, Jie Yu, Joseph Wu, Kevin
Klues, and Vinod Kone.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Supported mesos containerizer recover to be nested aware.
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/mesos/containerizer.hpp 16f9e3e92e90fe7f8a0ebd24e567800e1f285bc9

>   src/slave/containerizer/mesos/containerizer.cpp 144b0db501d40d4e0bba12672723616bedd76e7e

> 
> Diff: https://reviews.apache.org/r/52235/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Gilbert Song
> 
>


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