[Bug] Using lerd link/lerd init results in a database non blocking bug

[Legendary4226]

Hi, before I start describing my issue, I'd like to congrats you for the work you've made ! "Proxying" commands to containers to avoid having any dev dependency on host is an idea I had since ~two years but I was lacking in knowledges and time. A big thank's to you, Lerd is an awesome project!

So let's get to the point: I created a new Laravel project, followed the Lerd CLI prompts after running lerd init/link (don't remember which exactly) and it apparently did something in background (a config somewhere probably) that leads to :

> php artisan
[WARN] could not start mariadb-11: mariadb-11 is active but not yet ready: mariadb-11 did not become ready within 30s
Laravel Framework 13.14.0

Usage:
  command [options] [arguments]
...

This warn takes some time to appear (~30s) and it's apparently something that timeouts. Like I run php artisan and it takes 30s to execute (it hangs until the WARN appear then the command is executed).

Remark: Running php artisan from any other linked projects does not trigger the WARN thing.

I don't know what to give to help you debug or if the issue is clear. I can reproduce the bug with ease so ask me what you need ^^

Current fix: I used the website's GUI to link the folder and it worked (no WARN thing)

[geodro]

Thanks for the kind words, and for the really clear report, the detail about the GUI link working while the CLI one hung is exactly what pointed at the cause.

Here's what was happening. When you run a command like php artisan, lerd makes sure the site's database service is up and then waits for it to actually accept connections before handing off. That readiness check runs a tiny ping inside the container, and for the whole MySQL and MariaDB family it was using mysqladmin. The trouble is that the mariadb:11 image (which is the default version of the mariadb preset) dropped the old mysqladmin command and only ships mariadb-admin now, so the ping could never succeed. The check just kept retrying until it gave up at 30 seconds, every single time, which is the hang you saw. Other linked projects were fine because they were on mysql or sqlite, and the GUI link happened to not put mariadb into your config the way the CLI prompts did, so it sidestepped the broken probe rather than really fixing it.

The fix is to probe MariaDB with mariadb-admin and leave MySQL on mysqladmin, which is up in #479. I also fixed the same thing in the Sail import path while I was in there. It'll go out in the next release. Until then, if you want the hang gone immediately you can switch that project to mysql or sqlite, or rebuild from the branch.

Thanks again for taking the time to write it up.

[geodro]

Released!