Fisherworks / sys_stats_restful

A python based restful api that simply returns cpu temperature and disk space of linux

Sys Stats Restful

README 中文版

I just simply need a restful API that can tell the HA (Home Assistant) service about CPU temperature of the home router, which is running over KVM virtualization on a compact size SBC with only passive cooling. So the data collected comes directly from the host OS instead of the router (on vm).

So this is it, done in kinda less than 3 hours, not quite sophisticated or flexible though.

What can this offer

curl http://<hostname>:9090/stats/<data_type>

The data_type can be either du, temps, boot_time for now. Then it can returns the sensor data immediately.

curl http://192.168.12.34:9090/stats/temps

{
    "code": 0,
    "status": "success",
    "data": {
        "cpu_thermal": {
            "curr": 57.78,
            "crit": 100.0
        },
        "gpu_thermal": {
            "curr": 57.22,
            "crit": 95.0
        }
    }
}

How To Make This Work

1. Clone the repo, and set up python3 based virtual env.
2. Install the pip dependencies using the requirements.txt included.
3. Run the service with python entry.py.
4. Check if the service works well through http clients like curl http://<hostname>:9090/stats.
5. Optional - if it's all good, refer the sys_stats_restful.service to set up systemd service, then the service can survive after reboot.
6. Optional - integrate the service with HA by the to_be_placed_in_ha_configuration.yaml to show the data on HA dash.

Configuration of my HW/OS

1. An ARM based SBC (Asus Tinker Board 2s) with Armbian as host OS, kernel version 6.1.x #1 SMP PREEMPT aarch64 GNU/Linux.
2. OpenWrt Snapshot 23.x runs over KVM on the Armbian, which has 2 virtual eth interfaces bridged with 802.1q VLAN interfaces of Armbian.
3. The CPU and GPU temperature of the SBC can be viewed by sensors in terminal after apt install lm-sensors.
4. This service can be deployed either in Armbian (on bare metal hardware) or docker/lxc if you prefer, as long as the code can extract the necessary data from certain paths of host OS. I suppose a deployment in KVM based vm is NOT OK because the data can not be fetched in that level of hardware isolation.