Index by title

Container init process

Problem

To enable running multiple processes, containers require process/service management. This is normally provided by some kind of init task (e.g. from sysvinit).

There are Docker-compatible replacements for full-fledged init's. Unfortunately they require either custom init scripts or service configurations (https://wiki.gentoo.org/wiki/Comparison_of_init_systems). The process of migration from OS-provided OpenRC init scripts is time consuming and error prone.

Usage of system's default sysvinit is hampered by following shortcomings:

Solution

Nevertheless it is possible to use sysvinit inside Docker container as an init process. Required steps are following:

1. Change sysvinit to exit init process on hard shutdown (runlevel 0) with following patch

--- sysvinit-3.01/src/init.c    2021-12-13 20:21:26.000000000 +0100
+++ sysvinit-3.01/src/init.c    2022-04-18 01:21:47.966751774 +0200
@@ -2367,6 +2367,11 @@
      read_inittab();
      fail_cancel();
      setproctitle("init [%c]", (int)runlevel);
+
+      /*
+       * Exit on halt - causes Docker container to stop.
+       */
+      if (runlevel == '0') exit(0);
    }
  }
        Write_Runlevel_Log(runlevel);

On Gentoo it's enough to put this patch inside /etc/portage/patches/sys-apps/sysvinit/exit-on-halt.patch and reemerge sysvinit.

2. Change Docker signal for container termination to SIGINT and set appropriate action in inittab

Container's docker-compose.yml:

services:
  gentoo-base:
    ...
    stop_signal: SIGINT

/etc/inittab inside container - replace reboot action with shutdown:

# What to do at the "Three Finger Salute".
ca:12345:ctrlaltdel:/sbin/shutdown -h now

Generating Ruby+Rails documentation

cd /var/www/localhost/htdocs/rubydocs

git clone --branch v3_3_0 --depth 1 https://github.com/ruby/ruby.git
sdoc --all --exclude='./test/*' --exclude='./spec/*' --main=ruby/README.md --output=ruby-3.3.0 ruby

git clone --branch v7.1.2 --depth 1 https://github.com/rails/rails.git
sdoc --all --exclude='./*/test/*' --main=rails/README.md --output=rails-7.1.2 rails

sdoc-merge --op=ruby-3.3-rails-7.1 ruby-3.3.0 rails-7.1.2

rm -rf ruby/ ruby-3.3.0/ rails/ rails-7.1.2/

This is also available as Redmine HowTo

Keeping Redmine repository in sync with Github without dedicated plugin (Apache CGI + Github Webhook)

This is a solution in case you don't want to install additional plugins just to keep repository synchronised. It requires you to have Apache webserver with access to repository you are trying to sync. Apache has to support running CGI scripts.

Clone Github repository

Clone repository and make sure it is accessible by webserver:

mkdir /var/lib/redmine/repo
chown apache /var/lib/redmine/repo
su -u apache git -C /var/lib/redmine/repo clone https://github.com/username/repo_name.git

Enable WS for repository management in Redmine

Go to https://your.redmine.com/settings?tab=repositories and:

Prepare CGI script

Any script you run on your server will do. Below is an example of Bash script that pulls git repository and notifies Redmine to fetch changesets (substitute <repository-api-key> with your own):

#!/bin/sh
# Requires: jq

REPO_PATH='/var/lib/redmine/repo'

# Empty stdin, Apache issue https://bz.apache.org/bugzilla/show_bug.cgi?id=44782
REPO_NAME=$(cat <&0 | jq '.repository.name' | tr -cd 'A-Za-z0-9_-')
if [ -z "${REPO_NAME}" ]  || [ ! -d "${REPO_PATH}/${REPO_NAME}" ]; then
  echo "Status: 400 Bad Request" 
  echo "Content-Type: text/plain; charset=utf-8" 
  echo
  echo "project: unrecognized" 
  exit 0
fi

/usr/bin/git -C "${REPO_PATH}/${REPO_NAME}" pull -n -q
result1=$?

PROJECT_NAME=$(echo "${REPO_NAME}" | tr '_' '-')
/usr/bin/curl --max-time 60 -s "https://your.redmine.com/sys/fetch_changesets?id=${PROJECT_NAME}&key=<repository-api-key>" >/dev/null
result2=$?

if [[ $result1 && $result2 ]]; then
  echo "Status: 200 OK" 
else
  echo "Status: 500 Internal Server Error" 
fi

echo "Content-Type: text/plain; charset=utf-8" 
echo
echo "project: ${PROJECT_NAME}" 

if [[ $result1 ]]; then
  echo "git pull: ok" 
else
  echo "git pull: failed" 
fi

if [[ $result2 ]]; then
  echo "fetch changesets: ok" 
else
  echo "fetch changesets: failed" 
fi

Let's say you save this script under: /var/www/cgi-bin/update-repo.cgi

You can test if script executes properly:

echo <copy-input-from-github-webhook-request> | sudo -u apache /var/www/cgi-bin/update-repo.cgi

Configure Apache to run script whenever particular URL is requested

Inside VirtualHost of your choice just add:

  ...
  # Github webhook for repository pull/update
  ScriptAlias /update-repo.cgi /var/www/cgi-bin/update-repo.cgi
  <Directory /var/www/cgi-bin/>
    Options ExecCGI
    AllowOverride None
    Require all granted
  </Directory>
  ...

In case you use the same VirtualHost to proxy requests to your Redmine rails server, you should exclude your special URL from being proxied with:

ProxyPass /update-repo.cgi !

Configure Github Webhook

Go to your Github repository page, choose Settings -> Webhooks -> Add webhook. Then set:

Update webhook and you're done.

Wiki

Docker

Documentation

Github