minhhpham / cluster-resource-tracker

RESOURCE TRACKING FOR COMPUTER CLUSTERS

This project is created to serve users of computer clusters where ssh-ing to a compute node is not allowed. The tools in this repository will allow users to track resources in compute nodes without using ssh.

Currently, supported resources are:

• CPU usage (through Linux' ps command)
• Memory usage (through Linux' ps command)
• GPU usage (through nvidia-smi command)

HOW TO INSTALL AND SETUP

On Compute Node

Requirements: python and pip, assuming systemd is being used.

git clone https://github.com/minhhpham/Cluster-Resource-Tracker
cd Cluster-Resource-Tracker
pip install -r requirements.txt
python configure.py
chmod +x install-server.sh
sudo ./install-server.sh

If firewall is enabled, you will need to allow connection to port 2108. For example, with firewalld:

firewall-cmd --zone=trusted --permanent --add-port=2108/tcp
firewall-cmd --reload

On Control Node/Head Node

git clone https://github.com/minhhpham/Cluster-Resource-Tracker
cd Cluster-Resource-Tracker
chmod +x isntall-client.sh
sudo ./install-client.sh