Server stats for a remote server

[kubakubakuba]

Hello, is the server-stats widget available for remote servers? The https://github.com/glanceapp/agent link seems dead.

[kubakubakuba]

I have been able to solve this by writing a simple flask app:

import platform
import psutil
import time
from flask import Flask, jsonify, request, abort

app = Flask(__name__)

VALID_TOKEN = "your-secret-token"

def get_system_info():
    info = {
        "host_info_is_available": True,
        "boot_time": int(psutil.boot_time()),
        "hostname": platform.node(),
        "platform": platform.system(),
        "cpu": {
            "load_is_available": True,
            "load1_percent": int(psutil.getloadavg()[0] * 100),
            "load15_percent": int(psutil.getloadavg()[2] * 100),
            "temperature_is_available": True,
            "temperature_c": get_cpu_temperature()
        },
        "memory": {
            "memory_is_available": True,
            "total_mb": psutil.virtual_memory().total // (1024 * 1024),
            "used_mb": psutil.virtual_memory().used // (1024 * 1024),
            "used_percent": int(psutil.virtual_memory().percent),
            "swap_is_available": True,
            "swap_total_mb": psutil.swap_memory().total // (1024 * 1024),
            "swap_used_mb": psutil.swap_memory().used // (1024 * 1024),
            "swap_used_percent": int(psutil.swap_memory().percent)
        },
        "mountpoints": [
            {
                "path": partition.mountpoint,
                "name": partition.mountpoint,
                "total_mb": psutil.disk_usage(partition.mountpoint).total // (1024 * 1024),
                "used_mb": psutil.disk_usage(partition.mountpoint).used // (1024 * 1024),
                "used_percent": int(psutil.disk_usage(partition.mountpoint).percent)
            }
            for partition in psutil.disk_partitions()
            if not partition.mountpoint.startswith(("/snap", "/boot/efi"))
        ]
    }
    return info

def get_cpu_temperature():
    try:
        with open("/sys/class/thermal/thermal_zone0/temp", "r") as f:
            temp = int(f.read().strip()) / 1000.0
        return int(temp)
    except FileNotFoundError:
        return 0

@app.route('/api/sysinfo/all', methods=['GET'])
def sysinfo():
    token = request.headers.get('Authorization')
    if token != f"Bearer {VALID_TOKEN}":
        abort(401, description="Unauthorized: Invalid token")

    info = get_system_info()
    return jsonify(info)

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=8080)

You can then run it by flask run or python3 app.py, or you can set up a systemd service like:

[Unit]
Description=Glance server status
After=network.target

[Service]
User=ubuntu
Group=ubuntu
WorkingDirectory=/home/ubuntu/glance-serverstatus
ExecStart=/bin/python3 /home/ubuntu/glance-serverstatus/app.py
Restart=always

[Install]
WantedBy=multi-user.target

Then, you can use the serverstatus widget like normal:

- type: server-stats
          servers:
            - type: local
              name: Local
            - type: remote
              name: Remote
              url: http://remote.server.com:8080
              token: your-secret-token

I can package it inside of a Docker container for easier setup if need arises.

[kubakubakuba]

I packaged it inside a Docker container, you can set it up as:

git clone https://github.com/kubakubakuba/glance-serverstatus-agent.git
cd glance-serverstatus-agent
nano .env
docker compose build
docker compose up -d

[msalmasi]

Thanks for creating this.

This is working on the remote machine and I can pull stats manually using the /api/sysinfo/all endpoint. However, on the host running Glance, trying to use the Server Stats widget I'm getting this error trying to connect:

2025/03/10 17:12:51 WARN Getting remote system info: json: cannot unmarshal number 556 into Go struct field .cpu.load15_percent of type uint8

Any ideas how to resolve this?

EDIT:

I figured out the issue.

The raw values returned by psutil.getloadavg() are not percentages. They represent the actual number of processes contending for CPU time. To get a percentage, you need to divide the load average by the number of CPU cores. For example, on a 4-core system, a load average of 4.0 means 100% utilization, while a value of 8.0 indicates that the system is oversubscribed, with twice as many processes ready to run as there are cores.

To resolve this I changed the code as follows:

        "load1_percent": int((psutil.getloadavg()[0]/psutil.cpu_count()) * 100),
        "load15_percent": int((psutil.getloadavg()[2]/psutil.cpu_count()) * 100),

Now it works great!

[kubakubakuba]

Good job finding this! I did my testing on 1-core systems, so this did not show up. I will fix it in my repo, thanks!

[msalmasi]

This also has some issues dealing with volumes/mount points when Dockerized - it doesn't seem to show the correct data on Glance's dashboard unless I specifically point to the mount point "/app" like this

    "mountpoints": [
        {
            "path": partition.mountpoint,
            "name": partition.mountpoint,
            "total_mb": psutil.disk_usage('/app').total // (1024 * 1024),
            "used_mb": psutil.disk_usage('/app').used // (1024 * 1024),
            "used_percent": int(psutil.disk_usage('/app').percent)
        }
        for partition in psutil.disk_partitions()
        if not partition.mountpoint.startswith(("/snap", "/boot/efi"))

I also tried simply setting psutil.disk_partitions(all=True), which does send information about the mounted volumes but isn't correctly parsed by Glance. Only hard coding the mount point seems to work even though that is very janky. Someone else should be able to come up with a more robust solution that can properly account for volume mount points set in the docker-compose.

[kubakubakuba]

True, the mountpoints are wrong (at least the size seems okayish):

will try to look into it, but no idea if it is even possible.

[msalmasi]

It works when I specifically point to the mount point like this:

        "total_mb": psutil.disk_usage('/app').total // (1024 * 1024),
        "used_mb": psutil.disk_usage('/app').used // (1024 * 1024),
        "used_percent": int(psutil.disk_usage('/app').percent)

One way to handle this would be to use an environment variable here so that at least this can be set in the .env file rather than manually hardcoding the mount point in the app code. Then the use would have to specify the mount point he is using. There is probably a much more robust / automated solution though.

[kubakubakuba]

Maybe the cleanest solution would really be to explicitly define a list of mountpoints in the env file, I'll try it.

[Warren-0]

Thanks for creating this also.. ^ That fixed it for me too. I did have to change the sys../..thermal path for the correct cpu sensor for me. But that is expected.. Thanks again.

Example sensors on this i9

$ sensors
iwlwifi_1-virtual-0
Adapter: Virtual device
temp1: N/A

nvme-pci-0100
Adapter: PCI adapter
Composite: +41.9°C (low = -0.1°C, high = +76.8°C)
(crit = +78.8°C)

acpitz-acpi-0
Adapter: ACPI interface
temp1: +27.8°C

coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +49.0°C (high = +100.0°C, crit = +100.0°C)
Core 0: +42.0°C (high = +100.0°C, crit = +100.0°C)
Core 4: +44.0°C (high = +100.0°C, crit = +100.0°C)
Core 8: +44.0°C (high = +100.0°C, crit = +100.0°C)
Core 12: +46.0°C (high = +100.0°C, crit = +100.0°C)
Core 16: +44.0°C (high = +100.0°C, crit = +100.0°C)
Core 20: +43.0°C (high = +100.0°C, crit = +100.0°C)
Core 24: +46.0°C (high = +100.0°C, crit = +100.0°C)
Core 25: +46.0°C (high = +100.0°C, crit = +100.0°C)
Core 26: +45.0°C (high = +100.0°C, crit = +100.0°C)
Core 27: +45.0°C (high = +100.0°C, crit = +100.0°C)
Core 28: +48.0°C (high = +100.0°C, crit = +100.0°C)
Core 29: +48.0°C (high = +100.0°C, crit = +100.0°C)
Core 30: +48.0°C (high = +100.0°C, crit = +100.0°C)
Core 31: +48.0°C (high = +100.0°C, crit = +100.0°C)

nvme-pci-5800
Adapter: PCI adapter
Composite: +43.9°C (low = -0.1°C, high = +82.8°C)
(crit = +84.8°C)
Sensor 1: +43.9°C (low = -273.1°C, high = +65261.8°C)

[Warren-0]

Hi,

I had to make a change to the memory section for "used_mem", it seemed to be showing me the incorrect ram used. Inside or out of the docker.

I changed the below for me - from, to:

"used_mb": psutil.virtual_memory().used // (1024 * 1024),

to

"used_mb": psutil.virtual_memory().total // (1024 * 1024) - psutil.virtual_memory().available // (1024 * 1024),

That gave me the correct ram used. The psutil.virtual_memory().used did not show the correct ram used due to cache/buffers or so.. The "available" shows the true used value.

I ran this in a docker and via systemd too.
Just some findings I found for my system type.

Cheers!