cpu.rt_runtime_us behaviour is different in RHEL 7 Machine as compared to RHEL 8
Issue
- Differences in
cpu.rt_runtime_us
behavior betweenRHEL 7
andRHEL 8
. - Real-time scheduling time budget limit behavior differs between
RHEL 7 and RHEL 8
RHEL 7:
=======
# cat /sys/fs/cgroup/cpuacct/user.slice/user-0.slice/session-1.scope/cpu.rt_runtime_us
100000
# perf stat stress-ng -c 1 -t 10 --sched fifo --sched-prio 99 --taskset 2
...
10.007720112 seconds time elapsed
4.381507000 seconds user <=---- [ almost 4 times ]
0.002936000 seconds sys
RHEL 8:
=======
# cat /sys/fs/cgroup/cpuacct/user.slice/cpu.rt_runtime_us
100000
# perf stat stress-ng -c 1 -t 10 --sched fifo --sched-prio 99 --taskset 2
stress-ng: info: [1814] setting to a 10 second run per stressor
...
10.007775763 seconds time elapsed
1.000498000 seconds user << user time showing 1 sec as expected
0.006842000 seconds sys
- When setting
cpu.rt_runtime_us
globally and for auser.slice
, the time command is showing moreuser+sys
time inRHEL 7
and showing the value under limit inRHEL 8
for Real-time priority scheduling task.
Environment
- Red Hat Enterprise Linux
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.