nixos-rebuild switch produces warnings about systemd-udev-settle service

[bennofs]

Issue description

When running nixos-rebuild switch, sometimes systemd-udev-settle fails to start with an error message:

$ systemctl status systemd-udev-settle
● systemd-udev-settle.service - udev Wait for Complete Device Initialization
   Loaded: loaded (/nix/store/qzvya9kh5wh3bkcryd3g14xax48i1klc-systemd-231/example/systemd/system/systemd-udev-settle.s
   Active: failed (Result: exit-code) since Tue 2017-01-17 17:35:55 CET; 53min ago
     Docs: man:udev(7)
           man:systemd-udevd.service(8)
  Process: 3392 ExecStart=/nix/store/qzvya9kh5wh3bkcryd3g14xax48i1klc-systemd-231/bin/udevadm settle (code=exited, stat
 Main PID: 3392 (code=exited, status=1/FAILURE)

Jan 17 17:33:54 c-cube systemd[1]: Starting udev Wait for Complete Device Initialization...
Jan 17 17:35:55 c-cube systemd[1]: systemd-udev-settle.service: Main process exited, code=exited, status=1/FAILURE
Jan 17 17:35:55 c-cube systemd[1]: Failed to start udev Wait for Complete Device Initialization.
Jan 17 17:35:55 c-cube systemd[1]: systemd-udev-settle.service: Unit entered failed state.
Jan 17 17:35:55 c-cube systemd[1]: systemd-udev-settle.service: Failed with result 'exit-code'.

Steps to reproduce

Run nixos-rebuild switch

Technical details

• System: NixOS: 17.03.git.f673243aff (Gorilla)
• Nix version: nix-env (Nix) 1.11.5
• Nixpkgs version: 17.03.git.f673243aff

[Mic92]

I can confirm this issue. It also delays restarting of other units (probably dependent services) after upgrade.

I do:

$ systemctl stop systemd-udev-settle.service

as a workaround when this happens, so that the other services can restart faster.
It only happens on my physical machine. My v-server never had this problem.

[ocharles]

Can confirm this is still happens on 17.09 (just happened right now, and has been happening on every single switch).

[Mic92]

My workaround is to use:

systemd.service.systemd-udev-settle.serviceConfig.ExecStart = "${pkgs.coreutils}/bin/true";

[qrilka]

@Mic92 how do you specify that? I get:

error: The option `systemd.service' defined in `/etc/nixos/configuration.nix' does not exist.

[Mic92]

there was a typo:
UPDATE 2: One also need to make it a list to override!

- systemd.service.systemd-udev-settle.serviceConfig.ExecStart = "${pkgs.coreutils}/bin/true";
+ systemd.services.systemd-udev-settle.serviceConfig.ExecStart = ["" "${pkgs.coreutils}/bin/true"];

[chrissound]

The solution posted above #21954 (comment) no longer seems to work after upgrading to 18.09 - has anyone else solved this in a different way?

[chrissound]

Ran into this issue when I didn't have the EFI boot partition mounted while rebuilding. I think I was starting up an old nixos configuration. Seems to work fine now.

[Mic92]

@chrissound I forgot to make it a list. The empty string makes systemd first resets the option so it can be overriden afterwards.

[Mic92]

A potential fix is also here:
#53446

I want to test this option a bit for myself first.