Incorrect limits in dockerfile when cgroup v2 is used

[funbiscuit]

Currently script start.sh contains following code to automatically set memory limits:

if [ "$MEM_LIMIT_MB" = "" ]; then
	DOCKER_LIM_FILE="/sys/fs/cgroup/memory/memory.limit_in_bytes"

	if [ -e "${DOCKER_LIM_FILE}" ]; then
		MEM_LIMIT_MB=$(($(cat ${DOCKER_LIM_FILE})/1024/1024))
		echo "Using process mem limit of ${MEM_LIMIT_MB}MiB from ${DOCKER_LIM_FILE}"
	else
    	MEM_LIMIT_MB="1536"
    	echo "No process mem limit provided or found, defaulting to ${MEM_LIMIT_MB}MiB"
	fi
fi

But it doesn't work in my case in kubernetes, printing following message:

No process mem limit provided or found, defaulting to 1536MiB
MEM_LIMIT_MB=1536
Using default heap size of MIN(41% of MEM_LIMIT_MB, 640MiB) = 629MiB
HEAP_SIZE_MB=629

But there is a limit on this container (set by default by modelmesh controller) equal to 448Mi. This leads to OOMKilled because Java heap tries to grow to 629MiB which is greater than actual limit. I checked and actual limit is present at this path:

[app@modelmesh-serving-test-server-745dc7fb57-fxvxv mmesh]$ cat /sys/fs/cgroup/memory.max 
469762048

This is due to cgroup v2 that stores limits differently (at different path). Is it possible to change start.sh script so it works with both cgroup versions?

[ckadner]

Thanks @funbiscuit