To get started I’ve read the installation manual on the gitlabhq wiki page as well as serveral other blogs and resources in order to cope with the bugging Centos 5.5 problems and outdated packages. (I’ll admit I don’t like this OS very much…)
There are quite a lot of tutorials and manuals got get gitlabhq working but most of them are either not intended for CentOS or don’t provide help for certain pitfalls…Here I’ll provide a step by step explanation on how I did it.
My intended target configuration for the gitlabhq server is as follows:
- git 126.96.36.199
- ruby 1.9.2
- mysql 5.5.20
- redis 2.4.6
- apache 2
- mod_passenger 3.0.11
- gitlabhq 2.1 stable
I use mysql as database backend instead of the pre-configured sqlite3 setup gitlabhq uses by default. It’s just the better choice for a production environment. Preinstalled software
I expect you’ve got a Centos 5.5 System up and running with the following software preinstalled:
- git 1.7.x
- python 2.4
- mysql 5.5
Installation and Configuration
So let’s start with the installation and configuration process. If you follow the guide top-down you’ll have a running gitlabhq setup in notime!
We start by installing the overall prerequired CentOS and source packages.
Firstly install required CentOS packages like the openssh-server and other packages required for the building or installation processes.
sudo yum install -y git openssh-server curl-devel openssl-devel httpd-devel apr-devel apr-util-devel libzlib-ruby zlib-devel python-devel python-setuptools libicu-devel gcc-c++ gcc ruby-devel libxml2 libxml2-devel libxslt libxslt-devel
You have to compile the yaml library directly from source, there are no current packages available for CentOS.
wget http://pyyaml.org/download/libyaml/yaml-0.1.4.tar.gz tar xzvf yaml-0.1.4.tar.gz cd yaml-0.1.4 ./configure --prefix=/usr/local make && make install
It’s the same with the redis key-value store, compile it from source to get a current version, redis is used by gitlabhq.
wget http://redis.googlecode.com/files/redis-2.4.6.tar.gz tar xvfz redis-2.4.6.tar.gz cd redis-2.4.6 make && sudo make install
There are sed scripts available to customize the redis standard configuration, just execute the command line statements below. It enables the redis daemon process and outputs logs to /var/log/redis.log:
mkdir /etc/redis /var/lib/redis sed -e "s/^daemonize no$/daemonize yes/" -e "s/^dir \.\//dir \/var\/lib\/redis\//" -e "s/^loglevel debug$/loglevel notice/" -e "s/^logfile stdout$/logfile \/var\/log\/redis.log/" redis.conf > /etc/redis/redis.conf
Fortunately there is an init script for redis and Centos available, you can get it from github.com:
wget https://raw.github.com/gist/257849/9f1e627e0b7dbe68882fa2b7bdb1b2b263522004/redis-server sed -i "s/usr\/local\/sbin\/redis/usr\/local\/bin\/redis/" redis-server chmod u+x redis-server mv redis-server /etc/init.d /sbin/chkconfig --add redis-server /sbin/chkconfig --level 345 redis-server on /sbin/service redis-server start
** Testing Redis Installation**
Test the redis installation by opening a telnet connection and setting/reading a key value pair:
telnet 127.0.0.1 6379 set attitude:today "happy" get attitude:today
Installation Resque Post-Receive Git Hooks
If you’d like to use the Redis/Reque post-receive hooks with gitlab, the following command starts resque detached from the current terminal session:
nohup bundle exec rake environment resque:work QUEUE=* VVERBOSE=1 RAILS_ENV=production PIDFILE=tmp/pids/resque_worker_QUEUE.pid
Although I’m not using Sqlite3 in my configuration I’ll explain the installation process anyway in case you stick with Sqlite3 instead of MySQL for the database backend. As expected, the Centos 5.5 Sqlite package is outdated and has to be compiled from source:
wget http://www.sqlite.org/sqlite-autoconf-3070800.tar.gz tar -zxvf sqlite-autoconf-3070800.tar.gz cd sqlite-autoconf-3070800 ./configure --prefix=/opt/sqlite3 && make && make install
To finish the setup process the new shared libraries have to be copied to the libraries search path:
cp -p /opt/sqlite3/lib/libsqlite3.so.0 /usr/lib64 cp -p /opt/sqlite3/lib/libsqlite3.so.0 /usr/lib
As we know by now the way to go is to compile the software from source to get a current version, it’s no different for our beloved Ruby…
wget http://ftp.ruby-lang.org/pub/ruby/1.9/ruby-1.9.2-p290.tar.gz tar xzvf ruby-1.9.2-p290.tar.gz cd ruby-1.9.2-p290 ./configure --prefix=/usr/local --enable-shared --disable-install-doc --with-opt-dir=/usr/local/lib make && make install
Add the ruby binary path (/usr/local/bin) to the gobal paths:
echo 'pathmunge /usr/local/bin' > /etc/profile.d/ruby.sh chmod +x /etc/profile.d/ruby.sh
And update the gem system, just to be sure…
gem update --system
The gitolite installation manuals can be quite perplexing and complicated. But if you get down to it the installation process is quite straight forward:
Create a new user to be used by gitolite
useradd --shell /bin/bash --comment 'git version control' -m git
Create a corresponding group and add the newly created user to it
usermod -a -G git git
Generate an ssh-key for the newly created git user without a password!
su git ssh-keygen -t rsa
Copy the created key to the git user home directory replacing the default name with the username
cp .ssh/id_rsa.pub git.pub
Clone the gitolite repository
sudo -u git -H git clone git://github.com/gitlabhq/gitolite /home/git/gitolite
Install gitolite by using the earlier copied git public key for the initial setup
sudo -u git -H /home/git/gitolite/src/gl-system-install sudo -u git -H sh -c "PATH=/home/git/bin:$PATH; gl-setup ~/git.pub"
During the installation you’re prompted to edit the gitolite config file. Replace the repository umask value as described below
$REPO_UMASK = 0007;
As the last step fix the repository directory permissions and owner
sudo chmod -R g+rwX /home/git/repositories/ sudo chown -R git:git /home/git/repositories/
This step is mandatory in order to get the localhost entry in the ssh known_keys file made
To test the gitolite installation clone the gitolite admin repository on the server into the tmp folder.
You have to do this using the git user, only this user has its public key in gitolite installed at the moment.
git clone git@localhost:gitolite-admin.git /tmp/gitolite-admin
If everything is in order you can delete the cloned repository afterwards…
rm -rf /tmp/gitolite-admin
Wer’re finally getting to the gitlabhq installation. Let’s start with the prerequired python and ruby libraries:
easy_install pip sudo pip install pygments sudo gem install bundler
If you’re using Sqlite3 as db backend, install the sqlite3 gem with the following command
gem install sqlite3 -v '1.3.4' -- --with-sqlite3-dir=/opt/sqlite3
This is also a good chance to install the by gitlabhq required charlock holmes gem
sudo gem install charlock_holmes -v '0.6.8'
If you’re getting SSL certificate verification errors use the no verify global variable and try the gem install command again:
Now clone the gitlab repository and install the gems (as user git)
c d git clone -b stable git://github.com/gitlabhq/gitlabhq.git cd gitlabhq
If you’re using the preconfigured Sqlite3 backend you can skip this step and proceed directly with the bundle install command further below. But if you decide to use MySQL as database backend you’ll have to make some small changes to the gitlabhq configuration as well as initialize the mysql database and add a query user for gitlabhq. First, edit the config/database.yml file and configure a production environment similar to this:
production: adapter: mysql2 username: gitlab password: "your_db_password" host: localhost database: gitlab
Second, you have to add the mysql2 gem to the Gemfile in the gitlabhq folder (you can remove the sqlite3 gem dependency)
To be sure it works, install the mysql2 gem separately:
gem install mysql2
I expect you have a MySQL server up and running that’s why I’m only describing the initial database setup. Login into the mysql client using the root user
mysql -u root -p
Create a user and a database for gitlabhq
CREATE USER gitlab@localhost IDENTIFIED BY 'here_is_the_db_password'; CREATE DATABASE IF NOT EXISTS gitlab; GRANT ALL PRIVILEGES ON gitlab . * TO gitlab@localhost;
Install bundle, load tables and initial data
Now proceed with the finishing touch; install the bundles and execute the initial db setup
bundle install --without development test bundle exec rake db:setup RAILS_ENV=production bundle exec rake db:seed_fu RAILS_ENV=production
I’m against running gitlabhq using the rails provided server so I’ll be using passenger instead… Let’s assume you’re using a working apache2 installation with the module configurations stored under /etc/httpd/conf.d and a vhost setup with the vhost configurations stored under /etc/httpd/hosts.
Start by installing the passenger gem and passenger itself
gem install passenger passenger-install-apache2-module
Copy the generated loadmodule statements to a new module file and store it under /etc/httpd/conf.d/passenger.conf
LoadModule passenger_module /usr/local/lib/ruby/gems/1.9.1/gems/passenger-3.0.11/ext/apache2/mod_passenger.so PassengerRoot /usr/local/lib/ruby/gems/1.9.1/gems/passenger-3.0.11 PassengerRuby /usr/local/bin/ruby
Here’s an example of a working vhost configuration, you can copy paste and adjust it to your needs:
ServerName gitlabhq.yourhost.com DocumentRoot /home/git/gitlabhq/public AllowOverride All Options -MultiViews CustomLog logs/gitlab/access combined ErrorLog logs/gitlab/error ServerAdmin email@example.com RackBaseURI / RackEnv production
It’s extremely important to add the apache user to the git user group to prevent access problems:
usermod -a -G git apache
I’ve used the following web resources to help me with my installation:
I’ve used the following web resources to help me with my installation:
|Gitlabhq||Ubuntu linux installation tutorial|
|Gitolite (Gitlabhq)||Tutorial for gitolite installation on ubuntu linux|
|Gitolite Installation Tutorial||In depth installation tutorial for gitolite|
|Redis||Installation tutorial for redis on CentOS 5.6|
|Nokogiri||Installation tutorial for Nokogiri|
|Ruby 1.9.2 on CentOS 5.5||Installation tutorial for Ruby and Passenger on CentOS 5.5|
One sad thing about gitlabhq is the overall 500 error page. You won’t get detailed information on what went wrong it just says “there was a problem, check your config”.
The following thing is essential if you’re having problem like for example creating new repositories and using mod_passenger:
The user running passenger must have write access to the gitolite repository folder! If he doesn’t have the permissions – fail!
For general problems try deleting the lockfiles generated by gitlabhq in the system temp directory:
sudo rm -rf /tmp/gitlabhq-gitolite*
I’m using a slightly different setup in our company mainly because I’ve forked the gitlabhq project to make some company-related changes. In order to speed up development and deployment I’m using capistrano to deploy my gitlabhq fork to the server. I’m not posting information about it here because it’s a special setup and won’t be useful to you if you’re not actively making changes to gitlabhq.
So that’s it with this you should get gitlabhq working on your CentOS 5.5 system. If you’ve got problem or comments feel free to leave them here…