bonded link goes down on reconfigure

[Marlin Cremers]

systemd-networkd brings down member interfaces of a bond when they get reconfigured

Jan 31 15:43:46 unique-slug systemd-networkd[2430]: pn1: Bringing link down

This has a result that a server has complete downtime for a few seconds, this has also been reported upstream https://github.com/systemd/systemd/issues/31165.

A pull request has been opened to fix this https://github.com/systemd/systemd/pull/31172

[Lukas Märdian (slyon)]

Looks like this has been fixed in upstream systemd, so there's probably not a lot that we can do on the Netplan side.

[Nick Rosbrook (enr0n)]

The upstream bug is tagged for a stable backport, so we *should* get this soon-ish in noble.

[Marlin Cremers (marlinc)]

Would it be possible to backport this to 22.04 as well? We are currently using very hacky workarounds with Netplan.io and networkctl to prevent this bug from being triggered and causing downtime.

[Daniel 'f0o' Preussker (dpreussker)]

Bump this is also affecting us causing quite a lot of downtime just to add a vlan interface

[Alexander (zukimorimkato)]

Hello,

Another bump from me, as it affects our infrastructure quite heavily, as we are currently in process of transitioning to Jammy realease. And with more than thousand nodes being redeployed alreday, even small task like add new route causes unwanted complications.

I don't know if i did everything correct, but PSA patch for canonical systemd repo for jammy branch, which has been applied in mainstream of systemd, and has been linked abovr already. I hope it helps to fasten the process.

[Ubuntu Foundations Team Bug Bot (crichton)]

The attachment "jammy_netplan_systemd.patch" seems to be a patch. If it isn't, please remove the "patch" flag from the attachment, remove the "patch" tag, and if you are a member of the ~ubuntu-reviewers, unsubscribe the team.

[This is an automated message performed by a Launchpad user owned by ~brian-murray, for any issues please contact him.]

[Alexander (zukimorimkato)]

@enr0n Hi, I apologize for tagging you directly; as you very busy man, but I'd like to clarify if there is anything I can do to help on that matter? I.e. fill SRU or prepare a some kind of PR to offload it in some way.

[Nick Rosbrook (enr0n)]

I will look into this more when working on the next SRU for jammy. But this is not a trivial backport because (a) the patch does not apply, and (b) the link_down_now() function does not exist in v249 (was added in v251). It might be fine to just use link_down(), but this all needs to be tested when I have more time.

AFAICT, the patches should already be included in 24.04 since upstream backported them to v255.4.

[Nick Rosbrook (enr0n)]

Alexander - are you still experiencing this issue? Or have you upgraded to 24.04?

[Alexander (zukimorimkato)]

Hello Nick,

Thank you for your follow up on that matter.

So far, we are still experiencing this issue, because until we upgrade to 24.04 there is some time we have to use 22.04 for sure.

[Nick Rosbrook (enr0n)]

Okay, I have attempted to backport this patch to Jammy today. However, the upstream commit mentioned depends on another commit[1]. Unfortunately, I previously looked into backporting that commit in bug 2003250, but as I describe in the bug, it does not behave correctly when backported to Jammy. The same issue came up today when I tested all of these patches together. In particular, the bond is never fully configured on boot, and the subordinate device is never fully re-configured after `networkctl reconfigure`.

I do not have the time to investigate the root cause and write a proper backport for these patches, so this will not be prioritized for 22.04. Sorry :/

[1] https://github.com/systemd/systemd/pull/25162/commits/ebf95ef2fa915a7f0e2e747651391f07aa2fb180

[Alexander (zukimorimkato)]

I understand. Thank you for the time spent on that issue, we appreciate the effort.

Have a nice day =)

[Ondrej Kuchar (ondrej-kuchar)]

Hello All,

is it planned or possible to have the fix backported to Ubuntu 22.04 LTS ?

[Nick Rosbrook (enr0n)]

This may be fixed as a result of bug 2003250 being fixed.