From: Michael K. <li...@mk...> - 2020-05-24 22:10:18
|
> Am 24.05.2020 um 23:54 schrieb Michael Keuter <li...@mk...>: > > > >> Am 24.05.2020 um 23:38 schrieb Michael Keuter <li...@mk...>: >> >> >> >>> Am 24.05.2020 um 23:24 schrieb Lonnie Abelbeck <li...@lo...>: >>> >>> >>> >>>> On May 24, 2020, at 3:49 PM, Michael Keuter <li...@mk...> wrote: >>>> >>>> Try running the "yes" stresstest in the container, with the above limitations. >>> >>> If I define lxc.cgroup.cpuset.cpus to 1, or 2 or 3, the "yes" stress-test [1] only uses one core. >>> >>> Setting lxc.cgroup.cpuset.cpus=4 the container does not start. >>> >>> Not setting lxc.cgroup.cpuset.cpus all 4 cores are used. >>> >>> Lonnie >>> >>> [1] >>> (start test) >>> # for x in 1 2 3 4; do ( yes >/dev/null & ); done >>> >>> (stop test) >>> # killall yes >> >> ---- >> From the man page: >> >> lxc-cgroup -n foo cpuset.cpus "0,3" >> assign the processors 0 and 3 to the container. >> ---- >> >> I guess you define which CPU cores are allowed to use (0-3), that's why 4 does not work :-). >> >> lxc.cgroup.cpuset.cpus=0,3 (maybe in quotes, should use core 1 and 4) > > Yup, this uses CPU 2 + 4 (core 0 + 3), you define the actual cores NOT the count! > > lxc.cgroup.cpuset.cpus = 1,3 The RAM limit seems also be working fine. I set this in the container config: lxc.cgroup.memory.limit_in_bytes = 256M For the test you need to install the "pv" command to see the actual RAM value: ---- root@deb10:# dd if=/dev/zero bs=1024 | pv -b -B 1024 | dd of=/dev/shm/fill bs=1024 dd: error writing '/dev/shm/fill': Cannot allocate memory 240MiB 246721+0 records in 246720+0 records out 252641280 bytes (253 MB, 241 MiB) copied, 4.63373 s, 54.5 MB/s ---- Sometimes the container crashes at the end of that test. Michael http://www.mksolutions.info |