[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Custom Service Definition for Caddy

From: Efraim Flashner
Subject: Re: Custom Service Definition for Caddy
Date: Thu, 28 Jan 2021 20:17:07 +0200

On Tue, Jan 26, 2021 at 11:02:58AM -0600, Cameron wrote:
> Hello again everyone!
> I'm having a great time learning about Guix, but am struggling with defining 
> a custom service for Caddy. The 'caddy' package below works as expected -- 
> installing it puts the "caddy" binary in $PATH and it works fine when I run 
> it from a shell.
> #+begin_src 
<code snipped>
> #+end_src
> What doesn't work as expected is this service definition:
> #+begin_src 
> (define-record-type* <caddy-configuration>
>   caddy-configuration  make-caddy-configuration caddy-configuration?
>   (config-file          caddy-configuration-config-file
>                         (default "/etc/Caddyfile")))
> (define caddy-service-type
>   (shepherd-service-type
>    'caddy
>    (lambda (config)
>      (shepherd-service
>       (documentation "Run the caddy daemon (caddy).")
>       (provision '(caddy))
>       (requirement '(user-processes))
>       (start #~(make-forkexec-constructor
>                            (list "caddy" "run"
>                                  "-config" "/etc/Caddyfile")
>                                  #:log-file "/var/log/caddy.log"))
>       (stop #~(make-kill-destructor))))))
> #+end_src
> ...which I then add to my services list in config.scm like this:
> #+begin_src 
>                        (service caddy-service-type
>                                 (caddy-configuration
>                                  (config-file "/etc/Caddyfile")))
> #+end_src
> With this setup, 'herd start caddy' hangs for 30 seconds, during which time 
> the server is active and taking traffic according to the config in 
> /etc/Caddyfile but then something (presumably Shepherd) sends it a SIGTERM 
> and it dutifully shuts down. The 'herd start caddy' then exits with an error:
> #+begin_src 
> root@tindall ~# START=$(date +%s); herd start caddy; END=$(date +%s); echo 
> $(($END - START))
> Service caddy could not be started.
> herd: failed to start service caddy
> 30
> root@tindall ~# tail -n 1 /var/log/caddy.log 
> {"level":"info","ts":1611680035.1811087,"msg":"shutdown 
> done","signal":"SIGTERM"}
> root@tindall ~# 
> #+end_src
> 'caddy run' will keep caddy in the foreground, but I see the same behavior 
> when I use 'caddy start' (which forks another process for the daemon and 
> exits immediately) in the service definition instead.
> I suspect something is wrong with the 'start' procedure I've defined, but I'm 
> struggling to figure out what it is.
> What am I missing?
> -Cameron

Does caddy fork when you run "caddy run?" 'make-fork-constructor' forks
when it runs the code, so if caddy also forks then you have a
double-forked codepath and shepherd thinks the process is gone, so it
goes to kill it. If caddy doesn't have something like --foreground or
--no-daemonize then you might be looking at something like exec-command
from the manual¹.

Alternatively it might work if you can set a PID file.


Efraim Flashner   <>   אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted

Attachment: signature.asc
Description: PGP signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]