Container fails to start: "Error reading host's cpuset.cpus"

LXD
1

When launching a container the container builds but then fails to start. Syslog has an error about being unable to read cpusets:

Apr  8 15:31:50 pirx lxd.daemon[4001]: time="2022-04-08T15:31:50+02:00" level=error msg="Error reading host's cpuset.cpus"

I’ve tried to launch both a focal and an impish container with similar results.

Host is impish, running latest/stable: 5.0.0-c5bcb80

It seems this could be a cgroups incompatib. however I’m not sure why it would fail even when container and host are on the same release (impish)

Any pointers?

2

How does it fail to start exactly?

3

Pretty much silently. The lxc launch returns, I only see it being “Stopped” when doing lxc list

Same with lxc start.

The only indication of a fault is in the syslog

Some more context from syslog

Apr  8 13:57:47 pirx kernel: [18410.174073] lxdbr0: port 2(vethd753a130) entered blocking state
Apr  8 13:57:47 pirx kernel: [18410.174076] lxdbr0: port 2(vethd753a130) entered forwarding state
Apr  8 13:57:47 pirx systemd[3816]: snap.lxd.lxc.6b75d365-4f9b-4017-a0d6-a8d2977bfed2.scope: Deactivated successfully.
Apr  8 13:57:47 pirx systemd[3816]: Started snap.lxd.lxc.1c7a177e-88bc-48fa-a4fd-a8650fa79676.scope.
Apr  8 13:57:47 pirx kernel: [18410.305205] lxdbr0: port 2(vethd753a130) entered disabled state
Apr  8 13:57:47 pirx kernel: [18410.345068] lxdbr0: port 2(vethd753a130) entered blocking state
Apr  8 13:57:47 pirx kernel: [18410.345073] lxdbr0: port 2(vethd753a130) entered forwarding state
Apr  8 13:57:47 pirx kernel: [18410.365369] lxdbr0: port 2(vethd753a130) entered disabled state
Apr  8 13:57:47 pirx systemd-resolved[2269]: veth65f63a64: Bus client set search domain list to: lxd
Apr  8 13:57:47 pirx dnsmasq[4443]: using local addresses only for domain lxd
Apr  8 13:57:47 pirx dnsmasq[4443]: using local addresses only for domain lxd
Apr  8 13:57:47 pirx NetworkManager[2366]: <info>  [1649419067.7713] device (vethd753a130): released from master device lxdbr0
Apr  8 13:57:47 pirx systemd[3816]: snap.lxd.lxc.1c7a177e-88bc-48fa-a4fd-a8650fa79676.scope: Deactivated successfully.
Apr  8 13:57:48 pirx lxd.daemon[4001]: time="2022-04-08T13:57:48+02:00" level=error msg="Error reading host's cpuset.cpus"
Apr  8 13:57:48 pirx kernel: [18411.207777] audit: type=1400 audit(1649419068.567:214): apparmor="STATUS" operation="profile_remove" profile="unconfined" name="lxd-charmcraft_charmcraft-cloudsupport-35528151-0-0-amd64_</var/snap/lxd/common/lxd>" pid=113580 comm="apparmor_parser"
4

I am experiencing the same issue. Did you found any solution?

5

I haven’t I’m afraid

6

This may be what I’m running into as well. It just silently fails and lxd complains it can’t find the hosts cpuset. Ironically though, I have configured systemd to set that in the cgroups.

cat /sys/fs/cgroup/user.slice/user-1000.slice/cgroup.controllers                                                                                                                             
cpuset cpu io memory pids

They still do not start.