Doc: Better DNS persistence solution for systemd

[kkovacs]

Amended documentation with a better way to setup DNS-resolving LXD containers from the host machine.

This is a re-submit of #14572 .

[kkovacs]

Hi @simondeziel and @tomponline , I've re-submitted the PR as you asked.

I've added information regarding the situation Simon has encountered, and also opened #14588 , because lxdbr0 not being in a managed state is an issue that needs to be handled separately, not in this documentation snippet (even if I've added a workaround here).

[simondeziel]

@kkovacs If the existing .service hack doesn't work reliably on its own we might need to add something (udev rules?) to restart the service whenever the network device itself is changed.

[kkovacs]

I'm not very familiar with udev configuration. :) The .network file should be the correct solution IMHO; the one-shot service – even if somehow fortified by udev rules – is many more "rubber balls" on each other, and feels more hack-ish.

The .network solution is robust, and it would work immediately if the bridge was not out of sight for systemd-networkd after install. This temporary situation could easily cause other issues too, which are yet unknown.

But of course, it's up to the LXD project how to do this. I would be most happier if resolving the lxd-container-name.lxd names from the host would just work out of the box, and would not need a post-install snippet :)

[tomponline]

Please sign the CLA https://ubuntu.com/legal/contributors/agreement

[tomponline]

Please can you also rebase as main has fixes for some linter issues that are blocking the tests.

[kkovacs]

Please sign the CLA

I did!

Please can you also rebase as main has fixes for some linter issues that are blocking the tests.

I did :)

[tomponline]

Please can you sign off your commits.

See https://github.com/canonical/lxd/blob/main/CONTRIBUTING.md#developer-certificate-of-origin

[tomponline]

The CLA check is failing with

• user@kk-e1 ✕ (has no Launchpad account)

[kkovacs]

Signed off, removed "apparently", fixed "author", rebased, pushed. Fingers crossed. :)

[tomponline]

DCO and CLA check are good :)

[tomponline]

@minaelee happy with this?

[kkovacs]

Thank you @minaelee! It's incredible how much clearer you have made the text. Kudos!

[minaelee]

Suggested updates based on this conversation, thanks!

[minaelee]

lgtm!

[kkovacs]

Rebase, and hopefully commits will be signed now...

[tomponline]

Doc tests are failing with:

.tmp/doc/howto/network_bridge_resolved.md:113: MD009 Trailing spaces
.tmp/doc/howto/network_bridge_resolved.md:75: MD012 Multiple consecutive blank lines
.tmp/doc/howto/network_bridge_resolved.md:148: MD012 Multiple consecutive blank lines
.tmp/doc/howto/network_bridge_resolved.md:113: MD032 Lists should be surrounded by blank lines

Also please can you rebase.

[tomponline]

Please can you rebase this commit into your first commit so that we only have the final version of the logical change you're making, rather than commits showing the evolution of the PR. This is because we do not do squash merges and so the fully commit history of this PR will be merged into the main branch and we only want to see commits for logical changes.