CentOS Question posted in
The official documentation can be found here.

CentOS 8 Cgroup (or v2) installation without CWP

We are trying to install cgroup v2 to our centOS 8 virtual servers but after all configs we can limit memory bu we can’t limit cpu. We can’t see the cpu.max (sys/fs/cgroup/user.slice)

Has cgroup v2 got full support on CentOS 8 ? , we don’t have any idea on this. (We must use centOS 8 , can’t be Rhel 8)

Do you have any idea or solution about how cgroup V2 works on CentOS 8 without CWP ?

Thank You for All

Tags:

2

Answers


  1. 0

    ##https://www.redhat.com/en/authors/marc-richter
    #https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/pdf/managing_monitoring_and_updating_the_kernel/Red_Hat_Enterprise_Linux-8-Managing_monitoring_and_updating_the_kernel-en-US.pdf

    1. GRUB
    grubby --update-kernel=/boot/vmlinuz-$(uname -r) --args="cgroup_no_v1=all"
reboot
    1. /usr/bin/configure_cgroups.sh
    #!/bin/bash

touch /var/tmp/configure_cgroups.sh.ok

mkdir -p /mnt/cgroupv2

if [ ! -f /mnt/cgroupv2/cgroup.subtree_control ]
then
    mount -t cgroup2 none /mnt/cgroupv2
    echo "+cpu" > /mnt/cgroupv2/cgroup.subtree_control
    echo "+cpuset" > /mnt/cgroupv2/cgroup.subtree_control
    echo "+memory" > /mnt/cgroupv2/cgroup.subtree_control
fi

##cpuNo=0
numberfCpus=$(nproc --all)
numberOfUsers=$(ls /home| wc -l)
cpuShare=$((numberfCpus*100000/numberOfUsers))
totalMemory=$(cat /proc/meminfo | grep MemTotal| awk '{print $2}')
memoryShare=$((totalMemory/numberOfUsers*1024))
for username in $(ls /home)
do
  if [ ! -d /mnt/cgroupv2/user.$username ]
  then
      mkdir -p /mnt/cgroupv2/user.$username
      ##echo "$cpuNo" > /mnt/cgroupv2/user.$username/cpuset.cpus
      ##echo "$cpuNo" > /mnt/cgroupv2/user.$username/cpuset.mems
      echo "$cpuShare 100000" > /mnt/cgroupv2/user.$username/cpu.max
      echo "$memoryShare" > /mnt/cgroupv2/user.$username/memory.max
      ##cpuNo=$((cpuNo+1))
  fi

done

for username in $(ls /home)
do
      for pid in $(ps -ef |  grep -i "^$username" | awk '{print $2}')
      do
         grep $pid /mnt/cgroupv2/user.$username/cgroup.procs >/dev/null
         if [ $? != 0 ]
         then
             echo $pid > /mnt/cgroupv2/user.$username/cgroup.procs
         fi
      done
done
    1. SERVICE : /usr/lib/systemd/system/configure-cgroupsv2.service
    [Unit]
Description=Configure CGroups V2

[Service]
Type=oneshot
ExecStart=/usr/bin/configure_cgroups.sh
    1. SERVICE : /usr/lib/systemd/system/configure-cgroupsv2.timer
    [Unit]
Description=Configure CGroups V2 Timer

[Timer]
OnUnitActiveSec=10s
OnBootSec=10s

[Install]
WantedBy=timers.target

    systemctl daemon-reload
systemctl enable configure-cgroupsv2.timer
systemctl start configure-cgroupsv2.timer

systemctl list-timers --all| grep configure

systemctl start configure-cgroupsv2.service
systemctl status configure-cgroupsv2.service
journalctl -xe
    1. Testing :
    CPU: "while true; do   echo > /dev/null; done"  then  top
                                                   
RAM :  "cat /dev/zero | head -c 20000M | tail " then top
    Login or Signup to reply.
  2. 0

    Also , you must do the following to be able to run your docker images :

    1. yum install crun -y

2. cp  /usr/share/containers/containers.conf /etc/containers/

3. edit  /etc/containers/containers.conf 
   cgroups="disabled"
   runtime="crun"
    Login or Signup to reply.
Please signup or login to give your own answer.

Back To Top
Search