bug#46712: Removed service extensions are not unloaded from Shepherd upon reconfiguring

[Maxim Cournoyer]

Hello!

Ludovic Courtès <ludo@gnu.org> writes:

> Hi!
>
> At the Shepherd level, there’s no notion of service extension.
> Normally, after reconfiguring, “herd restart guix-daemon” (really
> “restart”, not “stop” + “start”) should start the new service, which
> doesn’t have all these ‘--chroot-directory’ options.

FTR, I had used 'guix deploy' and issued 'sudo herd restart guix-daemon'
on the remote after it completed successfully.  Why should stop + start
be different than restart though?  That seems counter-intuitive.

> Note that the guix-daemon process you should seems to be a child process
> (presumably because there was still a client running when you restarted
> the service), not the main guix-daemon process.
>
> You should check the command line of the main process, the one returned
> by “herd status guix-daemon”.

You are right that this process has the correct arguments:

$ sudo herd status guix-daemon
Status of guix-daemon:
  It is started.
  Running value is 25628.
  It is enabled.
  Provides (guix-daemon).
  Requires (user-processes).
  Conflicts with ().
  Will be respawned.

$ cat /proc/25628/cmdline
/gnu/store/rqif4yxa6ny4nxrdq6whnva2r089jm0c-guix-1.2.0-13.a53f711/bin/guix-daemon--build-users-groupguixbuild--max-silent-time0--timeout0--log-compressionnone--discover=no--substitute-urlshttps://ci.guix.gnu.org--max-jobs=20

Some of the other process were apparently caused by 'guix 
environment' shells still running in screen; I've terminated them all
now and ran 'sudo herd stop guix-daemon'; surprisingly I still had two
remaining guix-daemon processes that were launched manually for testing
purposes.  That's on a Guix system with an uptime of 174 days and
counting :-).

Thank you for the answer and sorry for the noise!  It works as designed.

Closing.

Maxim