resource.getrusage is showing different results inside docker container?
What I am trying to do:
- Running small scripts(c/c++/python3/java) inside a container in a file named temp.
- During the execution calculating the memory used by the script using getrusage(RUSAGE_CHILDREN).ru_maxrss
Solutions I have tried:
https://docs.python.org/3/library/resource.html#resource.RUSAGE_BOTH
Total memory used by Python process?
This is a simplified script I am using to run test.py and ./test inside a temp folder:
main.py
import subprocess
from resource import *
from datetime import datetime
import json
input_data = b"4\n5"
time = 5
status = True
mem_start = getrusage(RUSAGE_CHILDREN).ru_maxrss
print(mem_start)
try:
output = subprocess.run(f"cd temp/ && python3 test.py" ,shell=True , stdout=subprocess.PIPE, stderr=subprocess.PIPE , input=input_data , timeout=time)
result = output.stdout.decode()
print(result)
mem_end = getrusage(RUSAGE_CHILDREN).ru_maxrss-mem_start
except :
status = False
print("TLE")
mem_end = 0
status_data = {
"status" : "Successful" ,
"memory" : mem_end
}
print(status_data)
now if I run the below test.py script which allocates theoretical 78125kb of memory through the main.py script it will show 86544kb. Which is pretty close.
test.py
a = input()
b = input()
import sys
# allocating memory
a = [0]*10000000
print(sys.getsizeof(a)/1024)
But running a compiled c++ script it shows 5696kb(locally) and in the container 0kb which is not correct. The same thing happening for compile java script.
c++ script:
#include<iostream>
using namespace std;
int main()
{
int *p = new int(10000000);
int a , b ;
scanf("%d %d", &a , &b) ;
return 0;
}
Am I doing something wrong with python resources module? I have also used RUSAGE_SELF but it does not work.
I am running this script inside a nodejs server.
