Paulo Coghi

@e-dant Wow! Your project is a gem among the many watchers out there! Not only in regard to the OS APIs used, but to the code organization, excellent documentation, automated...

For our use case _(management of multiple servers through cli)_, tab-completion is a game-changer. Ex: on a command to show details of one server _(between hundreds)_ which is possible to...

> Pull requests are not accepted. Use the mailing list to submit patches.

I would like to note that the info on `/proc/loadavg` inside the container is correctly updated during the reproduction of this issue, and only `lxc-top` misses the real cpu usage...

I'm trying to understand what happens under the hood (on [src/lxc/tools/lxc_top.c](https://github.com/lxc/lxc/blob/main/src/lxc/tools/lxc_top.c)), in the hope to provide a PR with a fix, but I'm stuck in understanding the following things: -...

> The tool shouldn't look at the .monitor cgroup but at the .payload cgroup as that's the one which contains the actual tasks. I was the one who mistakenly assumed...

I'm not prepared enough yet to provide a fix. If you can, please go ahead :smiley: But I'will continue to study LXC source code, to be able to help and...

_Sorry for the long comment. I believe I'm on the right path but I apologize because what I will write here you probably already know for years_ :smile: ## cpu.stat...

> Possibly an issue with cgroup2. At least we can conclude that cpu.stat in cgroups v2 are working fine and reporting usage correctly on `/sys/fs/cgroup/lxc.payload./cpu.stat`.

> But if the data is originally provided by `/sys/fs/cgroup/lxc.payload./cpu.stat`, then the whole formula would need to consider everything based on microsecond instead of nanosecond. It seems to be the...