Glances Service Widget Incorrect Memory usage for Docker Containers

[gflip220]

Description

While using the Glances service widget (Containers metric), I have noticed the reported memory usage for Docker containers does not match what Glances and Docker Stats show. While Glances and Docker stats match, the Homepage widget shows sometimes wildly different memory statistics. This does not appear to be a timing issue as the numbers are consistently incorrect.

This screenshot, while cropped for size, was taken with all three windows on the screen at the same time. Note how several memory statistics on the widget are very far apart from Glances itself.

This is a freshly built Ubuntu 24.04 with latest versions of everything. Any thoughts as to what I may be able to check? Thank you!

homepage version

v0.10.9 (b5f4daa, Jan 6, 2025)

Installation method

Docker

Configuration

Basic glances/containers services.yaml entry:

   - Containers:          
        widget:
          type: glances
          url: http://IP:61208
          version: 4 # required only if running glances v4 or higher, defaults to 3
          metric: containers
          refreshInterval: 5000 # optional - in milliseconds, defaults to 1000 or more, depending on the metric
          pointsLimit: 15 # optional, defaults to 15

Container Logs

No response

Browser Logs

No response

Troubleshooting

Troubleshooting steps do not appear relevant. Logs are not showing any errors relative to this issue. Homepage and Glances widget appear to otherwise work correctly.

[shamoon]

homepage barely does anything to the glances api output which returns the values in bytes, you can look for yourself in your browser and do the conversion:

        "memory": {
            "usage": 63733760,
            "limit": 8214831104,
            "inactive_file": 1077248
        },

[gflip220]

Thanks for the reply. Everything appears to match up in the browser with Docker Stats. It's odd that some containers are just a little off by the time the data gets to Homepage, while others are massively incorrect (3GB for Plex in this example).

[shamoon]

Again, I would look at the api output, feel free to post it here

[gflip220]

Thanks for the clarification. I think I'm picking up what you are laying down now. Sorry, this isn't my area of expertise.

This is from the containers endpoint API output (http://IP:61208/api/4/containers) for the Plex container that is currently reporting 3GB usage. Is it correct to assume this means there is a problem with the Glances API? I guess the Glances web output doesn't pull from the same source as the API.

usage | 2962878464
limit | 33407393792
inactive_file | 2751602688 (Do you know what this value is?)

[shamoon]

Well yes, either a problem with glances or the container really is (was) using that much memory at the time. You may want to ask over in a glances venue

inactive_file is memory used by files (eg cached data) that are not actively in use but havent been freed up yet

[gflip220]

Thank you

[github-actions[bot]]

This discussion has been automatically closed because it was marked as answered. See our contributing guidelines for more details.

[shamoon]

FYI, turns out there is a bug here, inactive_file should be subtracted from the usage for accurate display. Not sure how I missed that in the glances docs.

#4844 will fix this