Workaround for cgroup - cpuset

[cyring]

CoreFreq driver is unaware of cpuset actions. Monitoring in the UI starts frozen.

For example with a cpuset "0-1,6-7"

Go to menu "HotPlug CPU" and disable all cpus which are not part of the cgroup cpuset

Monitoring should start afterward.

[cyring]

CPUSETS scripts

Sources:

• Reddit - Deprecated isolcpus workaround
• The Linux kernel user’s and administrator’s guide » Control Groups version 1 » CPUSETS
• ArchLinux cgroups

Create CPUSETS

CPUSET=/sys/fs/cgroup/cpuset
SETNAME=host
HOSTCPUS="0-1,6-7"

mkdir ${CPUSET}
mount -t cgroup -ocpuset cpuset ${CPUSET}

mkdir ${CPUSET}/${SETNAME}
/bin/echo ${HOSTCPUS} > ${CPUSET}/${SETNAME}/cpuset.cpus

Delete CPUSETS

rmdir ${CPUSET}/${SETNAME}

[cyring]

CPUSETS scripts

Sources:

• Reddit - Deprecated isolcpus workaround
• The Linux kernel user’s and administrator’s guide » Control Groups version 1 » CPUSETS
• ArchLinux cgroups

Create CPUSETS

CPUSET=/sys/fs/cgroup/cpuset
SETNAME=host
HOSTCPUS="0-1,6-7"

mkdir ${CPUSET}
mount -t cgroup -ocpuset cpuset ${CPUSET}

mkdir ${CPUSET}/${SETNAME}
/bin/echo ${HOSTCPUS} > ${CPUSET}/${SETNAME}/cpuset.cpus

Delete CPUSETS

sed -un p < ${CPUSET}/${SETNAME}/tasks > ${CPUSET}/tasks
rmdir ${CPUSET}/${SETNAME}
## Optionally
umount ${CPUSET}
rmdir ${CPUSET}

Issue

After deleting the CPUSETS and enabling the offline CPUs through the UI, two CPUs are still blocking monitoring when re-enabled.

EDIT Issue is gone, it was just a matter of time !

[cyring]

Need to add into README.md and Wiki, the Daemon and Client necessary cgroup binding.

[cyring]

Now part of the README.md