CODEWITHSUNDEEP
pythonpython-2.7docker
Anthony Kong
Anthony Kong

Reputation: 40624

How to correctly report available RAM within a docker container?

I have done the following test and I find that I cannot use psutil to retrieve meaningful data.

Basically I am trying to run this command to use python to report the available memory inside the container.

I have run the test three times and each time with a different memory parameter:

Here are the tests:

docker run -m 10000000  -ti python:2.7  bash  
docker run -m 100000000  -ti python:2.7  bash  
docker run -m 800000000  -ti python:2.7  bash

and here is the outcome

> docker run -m 10000000  -ti python:2.7  bash                                                                                                     
root@310de7b416cc:/# pip install psutil && python -c "import psutil; print psutil.virtual_memory().available"
Killed

> docker run -m 100000000  -ti python:2.7  bash                                                                                                    
root@e7aade23c143:/# pip install psutil && python -c "import psutil; print psutil.virtual_memory().available"
Collecting psutil
  Downloading psutil-5.3.1.tar.gz (397kB)
    100% |████████████████████████████████| 399kB 2.8MB/s 
Building wheels for collected packages: psutil
  Running setup.py bdist_wheel for psutil ... done
  Stored in directory: /root/.cache/pip/wheels/bc/00/68/b4cbf1017e55880ef2afd1a248a1c88311f38a574c8929d687
Successfully built psutil
Installing collected packages: psutil
Successfully installed psutil-5.3.1
1741164544

> docker run -m 800000000  -ti python:2.7  bash                                                                                                   
root@b8a28ad93432:/# pip install psutil && python -c "import psutil; print psutil.virtual_memory().available"
Collecting psutil
  Downloading psutil-5.3.1.tar.gz (397kB)
    100% |████████████████████████████████| 399kB 2.8MB/s 
Building wheels for collected packages: psutil
  Running setup.py bdist_wheel for psutil ... done
  Stored in directory: /root/.cache/pip/wheels/bc/00/68/b4cbf1017e55880ef2afd1a248a1c88311f38a574c8929d687
Successfully built psutil
Installing collected packages: psutil
Successfully installed psutil-5.3.1
1739603968

In first test, the job gets killed straight away. It is expected because of low RAM allocation.

But in second and third tests, even though the -m parameters are 8 times difference, the python code returns the more or less same results: 1741164544 and 1739603968 (i.e. 1,741,164,544 and 1,739,603,968)

Why it is the case?

My host PC is a mac and it has 16G RAM.

Upvotes: 4

Views: 3511

Answers (1)

myaut
myaut

Reputation: 11494

Docker uses CGroups to control resource usage by containers, so you should check CGroup files:

$ docker run -m 800000000  -ti centos  bash   
[root@c7406a25bc4b /]# cat /sys/fs/cgroup/memory/memory.limit_in_bytes
799997952

psutil is likely uses /proc/meminfo file which shows system's, not container, capabilities.

See also: https://serverfault.com/questions/680963/lxc-container-shows-hosts-full-ram-amount-and-cpu-count

Upvotes: 12

Related Questions