High CPU usage

[burgerga]

Check the bug
Previous issues were closed as solved

Describe the bug
Instead of <5% CPU Usage, I have around 15-25% CPU usage for glances on a Intel Core i3-8100T (running Proxmox)

To Reproduce
Steps to reproduce the behavior:

1. Install glances on Proxmox using https://raw.githubusercontent.com/nicolargo/glancesautoinstall/master/install.sh (via https://community-scripts.github.io/ProxmoxVE/scripts?id=glances)
2. Start via systemd service (/usr/local/bin/glances -w)
3. Open glances website
4. Observe high CPU

Expected behavior
Observe low CPU

Screenshots

Environement (please complete the following information)

• Operating System (lsb_release -a or OS name/version):
  Proxmox:

  Distributor ID: Debian
  Description:    Debian GNU/Linux 12 (bookworm)
  Release:        12
  Codename:       bookworm
  

• Glances & psutil versions:

  Glances version:        4.2.1
  Glances API version:    4
  PsUtil version:         5.9.4
  Log file:               /root/.local/share/glances/glances.log
  

• How do you install Glances (Pypi package, script, package manager, source): https://raw.githubusercontent.com/nicolargo/glancesautoinstall/master/install.sh
• Glances test:
  glances_issue.txt

Additional context
I have around 30 LXCs running in proxmox, so the processlist is quite long.
How can I further debug this?

You can also pastebin:

• the Glances configuration file (https://glances.readthedocs.io/en/latest/config.html#location): No config file
• the Glances log file (https://glances.readthedocs.io/en/latest/config.html#logging):
  glances.log

[burgerga]

Well, I guess this is indeed part of the problem:

processcount  [OK]    0.24868s
{'pid_max': 0, 'running': 1, 'sleeping': 922, 'thread': 2347, 'total': 1091}

But I also get high CPU if I request only CPU and MEM use using the gethomepage.dev glances info widget

[nicolargo]

Strange, Glances is now optimzed for CPU consumption (on my PC, it consume ~5% CPU).

Can you try this:

glances --disable-plugin all --enable-plugin quicklook,cpu,mem,swap,load,processlist

[burgerga]

Then it's between 8-15%, but mostly around 10%. If I remove processlist, and use top in a different shell I can see it does 3%.

[burgerga]

Apart from opening the website, I also tried using the API by using the homepage widget, which doesn't use the processlist endpoint (see https://github.com/gethomepage/homepage/blob/a06964dd175fa798fa9189ba1ecd5766bbde543f/src/pages/api/widgets/glances.js) but still causes high cpu... is this triggered by accessing the API?

[nicolargo]

You can also try to limit the number of processes to display in the WebUI.

In the glances.conf file:

[outputs]
...
# Limit the number of processes to display (in the WebUI)
max_processes_display=25
...

Yes for the moment, Glances update all plugins when a call is done to the API (but limited at one update every time). An optimisation can be discussed on this point but it is not straighforward because plugins can be linked (for example processlist and alert). I just open an issue to follow this specific point: #3033

[burgerga]

Hmm, that doesn't seem to matter very much, it's still around 20%.

However your linked issue is what I suspected. As a temporary workaround I now started a separate glances with:

glances --disable-plugin all --enable-plugin cpu,mem,fs,load,uptime,sensors -p 61209 -w --disable-webui

and connected it to my dashboard. It's now working perfectly with 0-3% CPU, and I can still access the cpu-hungry web interface at 61208 :)