CODEWITHSUNDEEP
linuxlinux-kernelcgroups
Judking
Judking

Reputation: 6371

Why my limit on CPU does not validate via cgroups

/etc/cgconfig.conf

...

group memlimit {
    memory {
        memory.limit_in_bytes = 8589934592;
    }
}

group cpulimit {
    cpu {
        cpu.shares = 1024;
    }
}

/etc/cgrules.conf:

@gatewayer  memory  memlimit/
@gatewayer  cpu cpulimit/

And I've restarted services via commands:

service cgconfig restart
service cgred restart

After I execute my python script from user who belongs to group gatewayer, I could see the PID when executing cat /cgroup/memory/memlimit/cgroup.procs and the limit have been applied on the running process. However, limit on cpu does not take effect, and cat /cgroup/cpu/cpulimit/cgroup.procs doesn't print the PID as expected.

I tried to check on the process, and it turns out to be the same result, that memory is limited by cgroups whereas cpu is not:

$ cat /proc/18113/cgroup 
174:blkio:/
173:net_cls:/
172:freezer:/
171:devices:/
170:memory:/memlimit
169:cpuacct:/
168:cpu:/
167:cpuset:/

Could anyone give me some help? Many thanks.

Upvotes: 0

Views: 1994

Answers (1)

Shuangistan
Shuangistan

Reputation: 1093

It seems the cgroups setting cpu.shares should be correctly applied.

Modern computer usually have a multi-core cpu. The python script used for testing can only use 100% of one core. So if there are still idle cpu cores, the other script could also use 100% from that core.

The better way to test cpu.shares is to run a number of processes more than the number of the cpu cores.

cat /proc/cpuinfo to indicate the number of cpu cores.

Upvotes: 1

Related Questions