You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by "Timothy St. Clair" <ts...@redhat.com> on 2014/07/08 18:57:21 UTC
Re: Review Request 22980: Enable support for existing cgroups,
don't always try to create.
> On June 25, 2014, 10:09 p.m., Ian Downes wrote:
> > src/slave/containerizer/isolators/cgroups/cpushare.cpp, line 236
> > <https://reviews.apache.org/r/22980/diff/1/?file=617510#file617510line236>
> >
> > Why would the cgroup already exist? The cgroup is based on the containerId which should be unique.
> >
> > Is this so the isolator can be re-used for controlling a cgroup added by something like Docker?
>
> Timothy St. Clair wrote:
> It's the mounted controller at the root.
>
> export MESOS_isolation='cgroups/cpu,cgroups/mem'
> export MESOS_cgroups_root='system.slice/mesos-slave.service'
> export MESOS_hierarchy=/sys/fs/cgroup
>
> If you try just starting, it will fail on the exists lines mentioned prior.
>
> Timothy St. Clair wrote:
> To elaborate -
>
> $ pwd
> /sys/fs/cgroup
>
> $ find . | grep mesos
> ./memory/system.slice/mesos-slave.service
> ...
>
> fails on 'memory'
>
> Timothy St. Clair wrote:
> & systemd creates the structure below prior to running the daemon.
>
> Vinod Kone wrote:
> I'm still confused. That exists() check is for the particular container's cgroup (e.g. /sys/fs/cgroup/memory/system.slice/mesos-slave.service/84a3b905-5ccc-4b4a-af07-ecf5be88fe53) which should not exist whether we are using systemd or not.
It cascades from the initial create->prepare calls. Jie was actually the person who helped point this out to me. Without the modifications slave simply falls over given the predefined input parameters.
- Timothy
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/22980/#review46691
-----------------------------------------------------------
On June 25, 2014, 7:52 p.m., Timothy St. Clair wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/22980/
> -----------------------------------------------------------
>
> (Updated June 25, 2014, 7:52 p.m.)
>
>
> Review request for mesos, Benjamin Hindman and Vinod Kone.
>
>
> Bugs: MESOS-1195
> https://issues.apache.org/jira/browse/MESOS-1195
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Currently if a controller is mounted outside of mesos perview it errors out. This fixes that issue.
>
>
> Diffs
> -----
>
> src/slave/containerizer/isolators/cgroups/cpushare.cpp 3265a80
> src/slave/containerizer/isolators/cgroups/mem.cpp e8d1e35
>
> Diff: https://reviews.apache.org/r/22980/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Timothy St. Clair
>
>