lxc-info 4.0.2 gives fewer information on pure cgroupv2

[emojifreak]

Host Linux: Ubuntu 20.04 with pure cgroupv2 (systemd.unified_cgroup_hierarchy=1)
Guest Linux: Ubuntu 19.10

LXC version:

# dpkg-query -W | grep lxc
liblxc-common	1:4.0.2-0ubuntu1
liblxc1	1:4.0.2-0ubuntu1
lxc-utils	1:4.0.2-0ubuntu1

In the Ubuntu default cgroup hierarchy (hybrid), lxc-info gives:

# lxc-info -n eoan-priv
Name:           eoan-priv
State:          RUNNING
PID:            2721
IP:             10.0.3.57
CPU use:        2.23 seconds
BlkIO use:      59.02 MiB
Memory use:     92.70 MiB
KMem use:       14.72 MiB
Link:           veth9qvFNx
 TX bytes:      1.63 KiB
 RX bytes:      8.07 KiB
 Total bytes:   9.70 KiB

On the other hand, when Host Linux is in pure cgroupv2, lxc-info only gives

# lxc-info -n eoan-priv
Name:           eoan-priv
State:          RUNNING
PID:            2825
IP:             10.0.3.57
Link:           vethcASHP0
 TX bytes:      2.08 KiB
 RX bytes:      9.52 KiB
 Total bytes:   11.60 KiB

This seems related to #3372

systemctl status gives almost as much information as lxc-info in the hybrid cgroup hierarchy as follows:

# systemctl status lxc@eoan-priv
● lxc@eoan-priv.service - LXC Container: eoan-priv
     Loaded: loaded (/lib/systemd/system/lxc@.service; disabled; vendor preset: enabled)
     Active: active (running) since Sun 2020-04-19 20:10:38 JST; 25s ago
       Docs: man:lxc-start
             man:lxc
   Main PID: 4226 (lxc-start)
         IP: 0B in, 696B out
         IO: 0B read, 16.0K written
      Tasks: 19 (limit: 19029)
     Memory: 32.0M
        CPU: 1.937s
     CGroup: /system.slice/system-lxc.slice/lxc@eoan-priv.service
             ├─lxc.monitor.eoan-priv
             │ └─4226 /usr/bin/lxc-start -F -n eoan-priv
             └─lxc.payload.eoan-priv
               ├─init.scope
               │ └─4243 /sbin/init
               └─system.slice

[weriyard]

I would to confirm this issue on Debian Bullseye and lxc 4.0.6

root@vdev1:/etc/apparmor.d/lxc# lxc-info one-706
Name: one-706
State: RUNNING
PID: 402463
IP: 192.168.184.2
Link: one-706-0
TX bytes: 1.94 KiB
RX bytes: 40.35 KiB
Total bytes: 42.29 KiB
Lack inforamtions about mem usage.

[graysky2]

According to the man page:

       -S, --stats
              Just print the container's statistics.  Note that for performance reasons the kernel does not  account  kernel
              memory use unless a kernel memory limit is set. If a limit is not set, lxc-info will display kernel memory use
              as 0. A limit can be set by specifying

                          lxc.cgroup.memory.kmem.limit_in_bytes = number

              in your container configuration file, see lxc.conf(5).

But even defining that key in my container config results in the lack of memory use:

# grep kmem /var/lib/lxc/p2/config
lxc.cgroup.memory.kmem.limit_in_bytes = 10000000

# lxc-info -n p2 -S
Link:           vethUozWOk
 TX bytes:      23.01 KiB
 RX bytes:      28.49 KiB
 Total bytes:   51.49 KiB

@brauner any thoughts?

EDIT: I tried lxc-info from the latest git but the results were the same.