CPU utilisation figure for Pod is not cumulative

[poVoq]

The large CPU utilisation figure for a Pod seems to display what looks like the highest current use by one of the included containers and not the cumulative CPU utilisation like it does for the RAM use of the Pod.

As CPU use by individual containers is already displayed below, it makes the Pod figure quite redundant, but a cumulative figure would be nice to have if the Pod includes many containers.

[jelly]

Showing the CPU / Memory usage was a bit tricky for pods so we use the following solution:

https://github.com/cockpit-project/cockpit-podman/blob/main/src/Containers.jsx#L488

So we just show the highest CPU usage of the shown containers as adding up CPU usage might show > 100% which might be weird/unexpected.

[poVoq]

IMHO, if it isn't cumulative, I would remove it completely as currently it is just redundant and misleading.

[martinpitt]

Please don't use line refs on main (as these change all the time). This is stable:

cockpit-podman/src/Containers.jsx

Line 486 in 3c5c63e

// As podman does not provide per pod memory/cpu statistics we do the following:

I agree to this request -- the CPU usage should just be added up. It's correct to be > 100% on multi-CPU systems -- top shows "%CPUs" as > 100% all the time.

If that is a bother, podman could also normalize the number by dividing by nproc. But then this would need to happen for the per-container metrics as well. The latter already show > 100% numbers:

but that is an independent request.