nixos/phpfpm: set systemd Type to notify-reload when possible

[aanderse]

Things done

as promised in #430490 (comment) adding notify-reload support

from php-fpm documentation:

systemd_interval int

When FPM is build with systemd integration, specify the interval, in second, between health report notification to systemd. Set to 0 to disable. Default value: 10.

i browsed through the php-fpm source and basically if this option isn't 0 then php-fpm will write READY=1\n to the systemd socket every n seconds as defined in config... which means incidentally notify-reload will work

---

btw... @roberth any idea what is going on here? did something change or was this error missed somehow in the last PR? when i run nix-build -A nixosTests.php.fpm-modular i get this error:

       error: The option `nodes.machine.system.services.php-fpm.meta' does not exist. Definition values:
       - In `/home/aaron/nixpkgs/pkgs/development/interpreters/php/service.nix':
           {
             maintainers = [
               {
                 email =

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

---

Add a 👍 reaction to pull requests you find important.

[roberth]

error: The option `nodes.machine.system.services.php-fpm.meta' does not exist.

Working on a better implementation, but #435005 will take care of the error in the meantime.

[Ma27]

Rest LGTM.

[roberth]

You might also be able to set WatchdogSec=, maybe derived from the php interval (times 2?), and/or vice versa.

(not a blocker)

[Ma27]

When I deploy this to a relatively normal Nextcloud installation, the reload runs into a timeout:

Sep 05 08:29:10 cloud php-fpm[1779427]: [NOTICE] Reloading in progress ...
Sep 05 08:29:10 cloud systemd[1]: Reloading PHP FastCGI Process Manager service for pool nextcloud...
Sep 05 08:29:10 cloud php-fpm[1779427]: [NOTICE] reloading: execvp("/nix/store/7c9pia7r4vvjb28lmndm46pkzqdc65wv-php-with-extensions-8.3.24/bin/php-fpm", {"/nix/store/7c9pia7r4vvjb28lmndm46pkzqdc65wv-php-with-extensions-8.3.24/bin/php-fpm", "-y", "/nix/store/c6lm1hzg6v4ygiinvw6s3inprp6ksxx1-phpfpm-nextcloud.conf", "-c", "/nix/store/xpq174hpbllkiw7hgryi43k91433ssxg-php.ini"})
Sep 05 08:29:10 cloud php-fpm[1779427]: [NOTICE] using inherited socket fd=7, "/run/phpfpm/nextcloud.sock"
Sep 05 08:29:10 cloud php-fpm[1779427]: [NOTICE] using inherited socket fd=7, "/run/phpfpm/nextcloud.sock"
Sep 05 08:29:10 cloud php-fpm[1779427]: [NOTICE] fpm is running, pid 1779427
Sep 05 08:29:11 cloud php-fpm[1779427]: [NOTICE] ready to handle connections
Sep 05 08:29:11 cloud php-fpm[1779427]: [NOTICE] systemd monitor interval set to 10000ms
[...]
Sep 05 08:29:56 cloud systemd[1]: phpfpm-nextcloud.service: Reload operation timed out. Killing reload process.
Sep 05 08:29:56 cloud systemd[1]: Reload failed for PHP FastCGI Process Manager service for pool nextcloud.

The reload signal was correctly handled given the logs, but apparently the sd_notify isn't happening?

Did you have some special configuration that made it work for you and we're probably missing in nixpkgs?

This is btw not a Nextcloud-specific thing, I can reproduce this with other phpfpm installations as well.

Blocking for now to prevent accidental merges until this is resolved.