Do you have any extra reads to get a better idea why you consider net_cls not a real cgroup?
Its generally used in iptables and tc traffic labeling an QoS.
Its a missing feature with an open issue that can be solved/eased in couple of lines of code https://github.com/systemd/systemd/issues/11618
Are PR welcome for this kind of feature?
On Wed, Nov 27, 2019 at 3:25 AM Lennart Poettering <lennart@xxxxxxxxxxxxxx> wrote:
On Di, 26.11.19 15:27, Andrew Gurinovich (altmind@xxxxxxxxx) wrote:
> What is the recommended way to assign cgroups prop net_cls.classid to a
> systemd service?
> Seems that a rule of thumb that whatever service property that is supposed
> to be systemd-managed, got a property in service file. However,
> https://systemd.io/CGROUP_DELEGATION says that:
>
> >>This means systemd currently does not and will never manage the following
> controllers on cgroup v1: freezer, cpuset, net_cls, perf_event, net_prio,
> hugetlb.
>
> Do It mean I'm on my own and need to write a custom ExecStartPost shell
> script to handle cgroups creation and cleanup myself just to change a
> single cgroup argument?
Yes.
These controllers are not supported by systemd, and this will not
change (at least for their cgroupsv1 implementations) even with
current systemd versions. On cgroupsv2, we now support "cpuset"
starting with the upcoming v244 of systemd, and "freezer" is currently
pending as PR. The other 4 do not exist and have no future on cgroupsv2.
Lennart
--
Lennart Poettering, Berlin
_______________________________________________ systemd-devel mailing list systemd-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/systemd-devel
