Script '/container/init/init.d' did not pass their completion check and container doesn't get healthy

[jvillegasd]

Summary

I was trying to deploy some new changes of my personal project into my VPS (I'm using kamal-deploy) and I got "Some initialization scripts haven't completed" error message. The weird thing is I didn't touch my Docker image in months. I setup env var DEBUG_MODE to TRUE to get logs.

Steps to reproduce

1. Create a simple Dockerfile with nfrastack/nginx-php-fpm:8.4-debian_trixie_faflopza as base image

2. enable the following PHP extensions:

ENV PHP_ENABLE_IGBINARY=TRUE
ENV PHP_ENABLE_MSGPACK=TRUE
ENV PHP_ENABLE_REDIS=TRUE
ENV PHP_ENABLE_OPENSSL=FALSE
ENV PHP_ENABLE_PDO_SQLITE=TRUE
ENV PHP_ENABLE_SQLITE3=TRUE
ENV PHP_MEMORY_LIMIT=256M
ENV DEBUG_MODE=TRUE

// Since we don't use zabbix remove it right away
ENV CONTAINER_ENABLE_MONITORING=FALSE

// Since this is production image, we don't need to create sample PHP files
ENV PHP_ENABLE_CREATE_SAMPLE_PHP=FALSE

// Since we are using Laravel workers, we don't need to enable cron
ENV CONTAINER_ENABLE_SCHEDULING=FALSE

ENV TIMEZONE=America/Bogota

RUN php-ext enable igbinary
&& php-ext enable msgpack
&& php-ext enable redis
&& php-ext enable posix
&& php-ext enable zip
&& php-ext enable pdo_sqlite
&& php-ext sqlite3

3. Copy the Nginx configuration
   COPY --chown=www-data:www-data ./config/some-nginx.conf /etc/nginx/sites.available/some-nginx.conf

4. Build

What is the expected correct behavior?

The docker image to build correctly.

Relevant logs and/or screenshots

Environment

• Image version / tag: nfrastack/nginx-php-fpm:8.4-debian_trixie_faflopza
• Host OS: Ubuntu 24.04

log.txt

[jvillegasd]

I see this ** Requested enabling 'session' module however it doesn't exist! in the logs but I'm not requesting this module to be enabled in my Dockerfile

[jvillegasd]

log-enhanced.txt I built and ran locally the dockerimage locally and I extracted enhanced logs

[tiredofit]

Lots of the environment variables have shifted from the previous "tireodfit" images..
For Example, you'll want to switch your extension commands from

ENV PHP_ENABLE_IGBINARY=TRUE
ENV PHP_ENABLE_MSGPACK=TRUE
ENV PHP_ENABLE_REDIS=TRUE
ENV PHP_ENABLE_OPENSSL=FALSE
ENV PHP_ENABLE_PDO_SQLITE=TRUE
ENV PHP_ENABLE_SQLITE3=TRUE

to

ENV PHP_MODULE_ENABLE_...=TRUE

The README also enables a bunch of modules by default, so you may want to switch the ones you dont want off. Some of the cross platform debian/alpine components don't always work, so I'll look into the session issue as it just causes confusion with stuff like that. I'm getting closer to finalizing a tagged release on this - Looking like 3rd week December before we're back to stability. There are lots of changes under the hood with php pool support, different runtime users and so on. We are getting closer.

Also. No need for CONTAINER_ENABLE_MONITORING=FALSE anymore, that's off by default. If you aren't using Zabbix or Fluent Bit or OpenBao you could probably switch to using the _core variants of the images, they are much smaller in size, but do the exact same things, just dont include the binaries.

I can't see where things are erroring out for some reason, Are you executing any scripts during initialization ?

https://github.com/nfrastack/container-bookstack/ is a really good image to see how I am building downstream images based on this image right now. Specifically how I'm loading extensions, setting them up to be used during the build so I can properly use composer, including nginx configuration (NGINX_SITE_ENABLED..) and the resulting file system layout that I use.

[jvillegasd]

Thanks for answering! I switched the env variables to the new ones and removed the monitoring setup. And yes, I am also executing 3 scripts related to Laravel by using CONTAINER_INIT_POST_COMMAND_PHP_FPM. I tried deploying without that env var but I got the same error message. I'm going to turn off the default modules I don't want as you mention

[tiredofit]

A way to bypass it could be CONTAINER_SKIP_SANITY_CHECK=TRUE but that defeats the purpose of the routines and won't help if other errors appear. Do you have this in a repo somewhere I can look at, even if private and get invited to it?

[jvillegasd]

I created a public repository with a minimum reproducible example: https://github.com/jvillegasd/php-docker-bug. It has a makefile for build and run the docker image. Please create a new .env file with the content of .env.example.

If you read the Dockerfile, you will notice I downgraded the image version from 8.4 to 8.3, that was me trying to test a different version.

I hope this can help you and I'm open for help, I will try to read the current source code of the Image for debugging

[tiredofit]

OK I'll see what I can find here.

This:

RUN php-ext enable igbinary \
    && php-ext enable msgpack \
    && php-ext enable redis \
    && php-ext enable posix \
    && php-ext enable zip \
    && php-ext enable pdo_sqlite \
    && php-ext sqlite3

Can be replaced by this, since you already have your ENVs set with PHP_MODULE..:

    php-ext prepare && \
    php-ext reset && \
    php-ext enable core && \

• That will, look for any new or upgraded modules or files from when the php image was first built and assign a priority tag to them on what order they are loaded (usually following php specifications 0,10,20). (prepare)
• (reset) will remove all loaded modules
• (enable core) will parse all the PHP_MODULE_*ENABLE=TRUE variables and perform symbolic linking, using something similar, or the php-ext tool above. I use those statements when I want to use composer in a Dockerfile, otherwise each container start after building will perform a reset/enable core each startup.

You have RUN chown -R www-data:www-data here, but if you are using the image as is there is no www-data user only a www-group. You can get away with using ${NGINX_USER} (nginx) and ${NGINX_GROUP} (www-data) or create your own users. They are set in the image in ENVs already.

The nginx configuration looks good, may need to be ported over to the new format. I believe I have pulled things out like the old snippets specifically this line:

    # Include additional security and optimization settings
    include /etc/nginx/snippets/site_optimization.conf;

However I don't see any init scripts in here that may be breaking it, it theoretically should just work here, unless we have some permissions issues. What I wrote above may solve it, otherwise I'll hve to look at your log files line by line again to see if there is anything. Also, you may want to turn off DEBUG_MODE which may reduce the noise and give us an error that may come up then we can search for it in the DEBUG statements...

[tiredofit]

OK, I am seeing the problem on a different image. It looks like I need to rebuild the base, nginx and nginx-php-fpm images after I build a fix.
Will let you know when new ones available.