Bad Gateway nginx for docker container - php

Question

I was trying to find out what is the problem for last 4 Hours, but had no luck.
I have two containers. PHP and nginx. First in docker/php/dockerFile:
FROM php:7.2.2-fpm
...
# Install Composer
...
# install node and npm
...
WORKDIR /var/www/
COPY post_run_web.sh /usr/local/bin/
RUN chmod 755 /usr/local/bin/post_run_web.sh
and the second one in docker/nginx/dockerFile:
FROM nginx:1.10
ADD ./vhost.conf /etc/nginx/conf.d/default.conf
WORKDIR /var/www
and in docker/nginx/vhost.conf
...
location ~ \.php$ {
try_files $uri /index.php =404;
fastcgi_pass app:9000;
fastcgi_index index.php;
fastcgi_buffers 16 16k;
fastcgi_buffer_size 32k;
fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
#fixes timeouts
fastcgi_read_timeout 600;
include fastcgi_params;
}
...
And my docker-compose.yml file
version: '3'
networks:
backend:
driver: bridge
frontend:
driver: bridge
services:
web:
build:
context: ./docker/nginx
dockerfile: dockerFile
container_name: "TEST_web"
volumes:
- ./:/var/www
ports:
- "80:80"
links:
- app
depends_on:
- app
networks:
- backend
app:
build:
context: ./docker/php
dockerfile: dockerFile
container_name: "TEST_php"
volumes:
- ./:/var/www
networks:
- backend
tty: true
entrypoint: ["/usr/local/bin/post_run_web.sh", "dev"]
...
I start my containers with docker-compose up. The problem is that my TEST_php container stoped automatically because of my post_run_web.sh in entrypoint option. So I added tail -f /dev/null in docker/php/post_run_web.sh to keep container running:
#!/bin/bash
cd /var/www
composer install
npm install
npm run $1
tail -f /dev/null
Now I see that all containers a running but get Bad Gateway nginx error when trying to access via browser. If I remove entrypoint from yml file, and try to execute post_run_web.sh manually after container started, everyting works fine.
How can I fix it and keep my entrypoint option?

Answer 1

I've been struggling with this the past few days. Some common problems I found are:
Not exposing the 9000 (or whatever other port php-fpm is listening on). This looks like your case.
php-fmp configuration: It can use a unix socket instead of listening in that
port. Look for in the php-fpm config. The listen directive should be listen = 0.0.0.0:9000. You may need to ADD/COPY the configuration.
The php-fpm service may not be running
The containers may not be linked (not your case)
I suspect that the problem may be that php-fpm is not started because you defined your own entrypoint, overriding the default php-fpm7.2 image's entrypoint which starts the service. Try starting the service in post_run_web.sh.
Hope this helps.