[Bug][6.1] : `lxc config device add <device_name> eth0 none` fails with ws error

[Gu1nness]

Required information

• Distribution: Ubuntu
• Distribution version: 24.04
• The output of "snap list --all lxd core20 core22 core24 snapd":

  Name    Version         Rev    Tracking       Publisher    Notes
  core20  20240416        2318   latest/stable  canonical**  base
  core22  20240408        1380   latest/stable  canonical**  base
  core24  20240528        423    latest/stable  canonical**  base
  lxd     6.1-90889b0     29398  latest/stable  canonical**  -
  lxd     5.21.1-2d13beb  28463  latest/stable  canonical**  disabled
  snapd   2.62            21465  latest/stable  canonical**  snapd,disabled
  snapd   2.63            21759  latest/stable  canonical**  snapd

• The output of "lxc info: https://zero.crans.org/?9030312503407366#DLiSrq41iyJEiU5pNABta7cnYpnE3p4R82pYjTncNKv6

Issue description

With lxc 6.1, running lxc config device add juju-e91cbe-25 eth0 none to simulate a network disruption used to fails (ret code 1) with

Error: websocket: close 1006 (abnormal closure): unexpected EOF

With lxc 5.21.1, the same command succeeds with Device eth0 added to juju-e91cbe-26

This has been confirmed by @shayancanonical

I tracked down a bit the error and it does not seem to fail on the client side:

• https://github.com/canonical/lxd/blob/main/lxc/config_device.go#L103
• https://github.com/canonical/lxd/blob/main/client/lxd_profiles.go#L64 So I would suppose an error on protocol server side.

This bugs happens since 6.1 got released in our CI:

• https://github.com/canonical/mysql-operator/actions/runs/9956496663/job/27507269588#step:22:376
• https://github.com/canonical/mongodb-operator/actions/runs/9941516387/job/27493445513?pr=433#step:24:170

Steps to reproduce

1. Deploy a working lxc container using juju : juju deploy mongodb-operator
2. Get the machine name with juju status
3. Simulate network cut: lxc config device add <machine_name> eth0 none

4. Information to attach

   • Output of lxd.buginfo : https://zero.crans.org/?ddd8c160c1c74299#FXK7gAU7UCHEYEjuy4xfMHeZ3CuW7E8Q6TV6wnfubd3i

[tomponline]

Thanks for the report we will look into this as a matter of urgency.

[tomponline]

In the future it would be great if you could run your CI using latest/candidate channel so we can catch issues like this before the release is pushed to stable :)

[Gu1nness]

In the future it would be great if you could run your CI using latest/candidate channel so we can catch issues like this before the release is pushed to stable :)

This is a really good idea and I'll implement that (maybe in a daily run) in order to catch this sooner!

[tomponline]

@Gu1nness https://github.com/canonical/lxd/pull/13775 will fix this.

We will cherry-pick into 5.21/candidate, latest/candidate and latest/stable shortly.

[tomponline]

For amd64 latest/candidate is fixed now.

[tomponline]

This is fixed in latest/stable now too.

[phvalguima]

hi @tomponline, running on ubuntu 24.04 and I can still see lxd v6.1 available: https://pastebin.ubuntu.com/p/WywCrWj4vv/

Same thing on my CI (22.04): https://github.com/canonical/opensearch-operator/actions/runs/9961773913/job/27544359496?pr=369

[tomponline]

This is fixed now in latest/stable 6.1-0d4d89b

[tomponline]

@phvalguima it would be great if you could also update your CI to use latest/candidate rather than latest/stable so that issues like this are picked up before new releases head to latest/stable. Thanks :)

[Gu1nness]

@tomponline We are discussing this right now with the Data Platform teams and will come with a solution soon!