exec driver CPU limit isn't honored when using cgroupv2

[valodzka]

Nomad version

Nomad v1.1.4 (acd3d78)

Operating system and Environment details

Debian GNU/Linux 11 with cgroups v2

Issue

CPU limit ignored when using driver=exec

Reproduction steps

Run tasks with different CPU resources and exec driver. See example job attached.

Expected Result

Tasks with a higher cpu limit get more cpu.

Actual Result

To difference in cpu usage, see screenshot (stress -v and stress -q have same cpu usage):

Compared with driver="docker". stress with -v gets much lower cpu percent:

Job file (if appropriate)

 group "stress" {
   count = 1
   task "stress" {

     driver = "exec"

     config {
       command = "stress"
       args =  ["--cpu", "12", "-q"]
     }
   #  driver = "docker"
   #  config {
   #    image = "progrium/stress"
   #    args = ["--cpu", "12", "-q"]
   #  }

     resources {
       memory = 10
       cpu = 1000
     }
   }
 }

 group "stress2" {
   count = 1
   task "stress" {
     driver = "exec"

    config {
       command = "stress"
       args =  ["--cpu", "12", "-v"]
     }
   #   driver = "docker"
   #  config {
   #    image = "progrium/stress"
   #    args = ["--cpu", "12", "-v"]
   #  }

     resources {
       memory = 10
       cpu = 10
     }
   }
 }

[lgfa29]

Hi @valodzka 👋

Thanks for the report. I was able to reproduce it when using cgroups v2. Using v1 it works, so it seems like we will need to update the task driver

[github-actions]

I'm going to lock this issue because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.