That makes sense. Then "71d35ca2-caa5-420d-9161-a42b750555cd" is not task id. Instead the task id is: "1408469503272-myuser-devel-largeprofile-0-d07b0417-eced-4515-b5ce-978416b82ebc".
If slave doesn't directly use task id or executor id, and instead use the random UUID for cgroup, then my assumption is that it maintains a mapping from this UUID to either task or executor id, internally. On Tue, Aug 19, 2014 at 11:28 AM, Vinod Kone <[email protected]> wrote: > Are you sure "71d35ca2-caa5-420d-9161-a42b750555cd" is your task id? The id > of the cgroup is a random UUID generated by mesos slave. It has nothing to > do with executor/task ids IIRC. > > > On Tue, Aug 19, 2014 at 11:22 AM, mohit soni <[email protected]> > wrote: >> >> I would like to know if cgroups are applied at executor level or >> directly at task level. Because, an executor can have many tasks, and >> executor can shrink or expand accordingly. Although, in most cases, >> only one executor runs one task. >> >> I ran a quick experiment, by launching an aurora job. When the task >> got launched on slave, here's the label that got created: >> /sys/fs/cgroup/cpu/mesos/71d35ca2-caa5-420d-9161-a42b750555cd >> >> The task got mounted under name: 71d35ca2-caa5-420d-9161-a42b750555cd >> >> I found that the executor id was: >> >> thermos-1408469503272-myuser-devel-largeprofile-0-d07b0417-eced-4515-b5ce-978416b82ebc >> >> And, the task id was: >> 71d35ca2-caa5-420d-9161-a42b750555cd >> >> Is this the expected behavior ? Or the cgroup label should be executor >> id ? Since, I am using aurora, thermos_executor is used for tasks. >> >> Mesos version: 0.18.0 >> >> Regards >> Mohit > >
