How to properly create cgroup with bash script?

Question

I'm trying to write init_cgroup.sh script, that will create container with specified in args limitations. As I understood, I need to find out the cgroup version at first, mount it, and then apply options to corresponding files

Here is how I tried to do it:

#!/bin/sh

# args:
# $1 - mem_limit (not used at the moment)
# $2 - core usage percentage (e.g. 0.1)
# $3 - container name
GROUP_ID=$3
GROUP_PATH="/sys/fs/cgroup"

mount -t tmpfs -o rw,nosuid,nodev,noexec,relatime,size=0k cgroup_root $GROUP_PATH
mount -t cgroup2 none $GROUP_PATH
if [ $? != "0" ]; then
    mkdir "$GROUP_PATH/$GROUP_ID"
    mount -t cgroup -o cpu,cpuset,cpuacct,memory cgroup_cpu "$GROUP_PATH/$GROUP_ID"
    echo 100000 > "$GROUP_PATH/$GROUP_ID/cpu.cfs_quota_us"
    echo "100000*$2" > "$GROUP_PATH/$GROUP_ID/cpu.cpu.cfs_period_us"
    cpus_count=$(grep -c processor /proc/cpuinfo)
    rand_cpu=$RANDOM%cpus_count
    echo $rand_cpu > "$GROUP_PATH/$GROUP_ID/cpuset.cpus"
else 
    {
    echo "+cpu" 
    echo "+cpuset" 
    echo "+memory"
    } >> "$GROUP_PATH/cgroup.subtree_control"
    mkdir "$GROUP_PATH/$GROUP_ID"
    cpus_count=$(grep -c processor /proc/cpuinfo)
    rand_cpu=$RANDOM%$cpus_count
    echo "$rand_cpu" > "$GROUP_PATH/$GROUP_ID/cpuset.cpus"
fi

P.S.: I should get as a result that the tasks in my group were executed on only one core, so I assigned it to a random cpu (temporary solution)

When running the script with sudo, it creates folders in

/sys/fs/cgroup, but outputs an error:

> sudo ./init_cgroup.sh 0 0.1 testt


> ./init_cgroup.sh: 29: echo: echo: I/O error

I figured out that I/O error causes because it hasn't have enough rights to write in system folder, even if I run it with sudo.

What should I do with it, and did I do it properly at all?

Answer 1

As @Cyrus mentioned sh is not an bash, so the first one isn't containing $RANDOM variable, and echo tried to write some trash in file. The solution was just change #!/bin/sh to #!/bin/bash