On Nov 25, 2013, at 6:59 PM, Kay Sievers <k...@vrfy.org> wrote: > On Mon, Nov 25, 2013 at 5:28 PM, Umut Tezduyar Lindskog > <umut.tezdu...@axis.com> wrote: >> On Nov 24, 2013, at 4:39 PM, Kay Sievers <k...@vrfy.org> wrote: >>> On Sun, Nov 24, 2013 at 12:33 PM, Umut Tezduyar Lindskog >>> <umut.tezdu...@axis.com> wrote: >>>> How do we support thread level resource management with the new cgroup >>>> abstraction? >> How do we do it in the process level then. Lets say a service has 5 >> processes under and 1 of them needs to be in a different slice. Any example? Can someone explain the process level management? >>>> >>>> Can we use scopes with task ids of threads? If so, what is the API to put >>>> the task id into its own scope unit? >>> >>> There is no plan at the moment to support thread-granular resource >>> settings. The feature is expected to be removed from the kernel too, >>> when we switch to the unified cgroup hierarchy. >> Any plans to support existing applications that are making use of thread >> level resource management? > > No, the current idea is that there will be no replacement. > >> If not, what are we left with then, posix thread priorities? > > Something else than cgroups, yes. Thread prios could work if it fits > the way your tool works. And there are some ideas of adding new > interfaces to /proc/$PID/, to provide these thread-level knobs. > > We will see when we get there. Only one thing seems clear, threads are > to be managed by the process, not by cgroups. > >> Is there an announcement of dropping thread level resource management >> support from kernel somewhere? > > Nothing official, there are just discussions about the "unified hierarchy". > > Kay
_______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel