> On Sept. 22, 2014, 7:12 p.m., Ben Mahler wrote:
> > Why do you need this?
> >
> > Seems like one would only watch for cgroup emptiness when destroying a
> > cgroup, in which case, why split emptiness waiting from a successful
> > destroy?
>
> Ian Downes wrote:
> This is for use in conjunction with pid namespaces: we kill the leading
> ('init') process and then we can wait until all processes have been killed by
> the kernel. The LinuxLauncher continues to use the freezer cgroup for this
> purpose (and for forward/backward compatibility).
Ok, 'emptied' is definitely a better name.
Seems unfortunate that we're pushing "watcher" logic into the cgroups library.
It's not obvious why the empty() watching logic couldn't be placed alongside
the code that is responsible for destruction of cgroups (including pid
namespaces), could you add a TODO to reflect future cleanup?
- Ben
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/25864/#review54170
-----------------------------------------------------------
On Sept. 23, 2014, 11:39 p.m., Ian Downes wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/25864/
> -----------------------------------------------------------
>
> (Updated Sept. 23, 2014, 11:39 p.m.)
>
>
> Review request for mesos, Jie Yu and Vinod Kone.
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Polls cgroups.procs until no processes in the cgroup. Poll interval and
> timeout can be specified.
>
>
> Diffs
> -----
>
> src/linux/cgroups.hpp abf31df1b4dbf6f715f93256b83c9996a45099cf
> src/linux/cgroups.cpp 62df4b7645c6ab061a47634058d79ca849caa6b9
>
> Diff: https://reviews.apache.org/r/25864/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Ian Downes
>
>
