[
https://issues.apache.org/jira/browse/MESOS-6265?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15530306#comment-15530306
]
haosdent commented on MESOS-6265:
---------------------------------
+1 For this. And we didn't use {{/sys/fs/cgroup/freezer/mesos/<x>/<y>/<z>}}
before 1.1.0, right?
> Adjust cgroups layout for nested containers.
> --------------------------------------------
>
> Key: MESOS-6265
> URL: https://issues.apache.org/jira/browse/MESOS-6265
> Project: Mesos
> Issue Type: Task
> Reporter: Jie Yu
>
> Currently, the freezer cgroup layout is:
> /sys/fs/cgroup/freezer/mesos/<x>/<y>/<z>
> I am wondering if we should change it to:
> /sys/fs/cgroup/freezer/mesos/<x>/mesos/<y>/mesos/<z>
> Two reasons:
> 1) Allow us to deal with user created cgroups. Imagine we start to support
> cgroups namespace and run docker daemon inside our container. We should allow
> docker to create cgroups underneath our cgroup, and we should not treat
> Docker cgroups as part of a nested container.
> 2) It's always nice to have a top level root for each level of nesting. That
> allows us to do more resource control like setting the memory limit for all
> nested containers in a level, or setting cfs quota for all nested containers,
> etc. in the future.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
