Wrong boot time in lxc containers

[ser-evegenii]

Telegraf system input plugin use btime in /proc/stat to count uptime. This behavior a little wrong, because in lxc containers btime show boot hoster time, but not container.
Maybe better get uptime metric in /proc/uptime

[danielnelson]

Can you raise this with the gopsutil project and link to the issue? Here is the function that is called for uptime: https://github.com/shirou/gopsutil/blob/master/host/host_linux.go#L93

[ser-evegenii]

shirou/gopsutil#390

[danielnelson]

This may have been fixed for a period, but doesn't seem to be working with LXC 3.
shirou/gopsutil#679

[jessingrass]

seems shirou/gopsutil#894 addresses the upstream issue.

[zeus86]

As far as i can tell - even if this issue is addressed upstream - if does not have any effect onto telegraf yet. LXC Containers still report the uptime of their hypervisor, instead of their own; at least on pxoxmox 6 and 7 (ubuntu focal guests, x86_64), where I tested this, so this should be repoenend. maybe a switch in the config-section from where to read the uptime would be suitable?

$ telegraf --version
Telegraf 1.20.2 (git: HEAD f721f53d)

[Astaoth]

Hi, I can confirm than Telegraf in a LXC on Proxmox 7 doesn't report the right uptime (Telegraf 1.22, VM and Proxmox up-to-date). The idea of @zeus86 seems fine, would it be possible to add this switch ?