nebula: Add restartTrigger for config file

[nh2]

Description of changes

Upstreaming this from my production setup.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 23.11 Release Notes (or backporting 23.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

CC nebula package maintainers @Br1ght0ne @numinit

[jian-lin]

I am not sure what "Current command vanished from the unit file" means and how it affects service restarting.

However, I do not think adding restartTriggers makes any difference. All restartTriggers does is adding X-Restart-Triggers = your-triggers to the service file.

• Since this directive starts with X-, systemd ignores it.
• Also, the logic of restarting services in switch-to-configuration.pl does not do anything special for it. And configFile has already been used in ExecStart. So there is no difference after adding restartTriggers from the perspective of switch-to-configuration.pl.

[nh2]

I am not sure what "Current command vanished from the unit file" means and how it affects service restarting.

It is an error/issue with systemd, see e.g. #49415. Sometimes, when a systemd service's unit file changes (e.g. when the command line changes as part of a NixOS configuration switch), there is a race where systemd notices that change, and then seems to run some "somebody changed my file under me, I'll do nothing about it" logic. Unfortunately it isn't documented in systemd, thus my vague explanation.

The effect of that is that sometimes on NixOS configuration switches, services aren't restarted (#49415), and in this case for Nebula, I noticed that when I change the Nebula config file via configuration.nix, Nebula was not restarted (thus not observing the config change).

• the logic of restarting services in switch-to-configuration.pl does not do anything special for it. And configFile has already been used in ExecStart. So there is no difference after adding restartTriggers from the perspective of switch-to-configuration.pl.

Hm, this is correct. From b606165 the only thing X-Restart-Triggers does is to be present in the unit file, so that the unit file changes. So you're right that should already tbe the case due to ExecStart = "${netCfg.package}/bin/nebula -config ${configFile}";.

I am wondering now whether I was just lucky regarding the race and this change indeed had no additional effect on my systems, or whether the fact that a string in the unit file changes that's outside the ExecStart line ("current command ...") helps systemd avoid the problem somehow.

I'm inclined to close this PR, until we have some proof that the change actually does something.

[nh2]

From my comment:

I'm inclined to close this PR, until we have some proof that the change actually does something.