FR: please add temperature metrics

[AtmanActive]

Hi, thank you for this wonderful software.

It could be even more useful in homelab environments if it could track temperatures.

Obviously, the most important one being CPU Package Temperature. Also useful could be Motherboard Sensor Temperature, and, finally Disk Drive Temperature (especially important for NVMes).

Thanks.

[c0m4r]

Thank you! I'll definitely add it.

[c0m4r]

Initial implementation of CPU thermal monitoring: ed0753f

[c0m4r]

Further enhancements for AMD: 6ce6aa4

Now I'll test it on different hardware and maybe some SBCs to verify. For sure this this gonna need some more work.

[c0m4r]

Intel N100

[AtmanActive]

Great progress, thanks!
Let me know when you have a new release and I'll give it a spin on machines I have access to.

[c0m4r]

New release is ready. Give it a try :)

[AtmanActive]

The metric seems to be working correctly, CPU only for now, but the graph is missing a defined max Y to make any sense. The correct value for max-Y for temperature should be TjMax, if available, otherwise it should be 100°C.
The way it is now, where the graph scales max-Y based on current data flow, is, sorry to say, useless :)

Thanks.

[c0m4r]

IMHO that's not entirely true. It may not be useful to you, but that's purely a matter of personal preference. In my experience, in most cases, temperature sensor data is presented without setting a maximum value. Does that mean you're wrong? No, they're simply two different ways of presenting data. There's no single correct way. And this is just one of many aspects of presenting data in charts. That's why many monitoring tools have so many toggles and configuration options, so everyone can tailor the data presentation to their preferences. I just want to avoid too many customization options, so that the dashboard remains relatively simple and doesn't require hours of learning the interface. That being said, I think I'll add the option to set a maximum value because it will definitely be useful. I'm thinking of three modes: off, on, and auto. In auto mode, we'll attempt to determine TjMax, and if that fails, we'll fallback to the default, say 100°C.

[AtmanActive]

I agree with you completely - to each their own.

It's just that if I look at a temperature graph, or a CPU usage graph, or RAM usage graph, then, without reading the values shown on the Y axis, I expect to be able to tell how much is utilized, despite the graph wiggling up and down. In that sense, for me personally, graphs that auto-scale are useless. That's because my point of view is utilization at a glance. If the graph is full but Y axis is showing, for example, 60% as max, then, I can't tell the utilization right away. I need extra brain cycles and focusing on Y axis markings to deduce what the current utilization is. In that case, I would rather not have a graph at all, than to have something misleading or hard to read.
Just my two cents.

Thanks.

[c0m4r]

Graph Y-axis max bound settings: bdf3d02

config:

  # Optional bounds limits for dynamic charts. "max_mode" can be:
  # "off": Unrestricted, Chart.js auto-scales Y-axis by stream data peaks.
  # "on": Imposes a hard Y-axis maximum matching "max_value". 
  # "auto": Tries to securely detect hardware constraints (e.g CPU TjMax or Net link speed). If failed, falls back to "max_value".
  graphs:
    cpu_temp:
      max_mode: "off"
      max_value: 100 # Celsius
    network:
      max_mode: "off"
      max_value: 1000 # Mbps (1 Gb/s)

[c0m4r]

Disk drives temperature monitoring: 8454ead
Also the dashboard is now split into two sections. The second one is "Thermals" and I put CPU and disk drives temperatures there.

[c0m4r]

New release is ready, try it out when you get the chance and let me know what you think :)

[c0m4r]

I noticed that auto-detect with some nvme drives results in weird values with decimal points like 93.07. I'll investigate it.