What is gitzebo?
gitzebo is a small Python git management web application. It was named gitzebo as it is intended to be a smaller, more lightweight, more secluded variant of the service GitHub and GitLab provide.
It is ideal for creating, managing, and sharing git repositories among small groups of developers.
I'm still using gitzebo myself. However, I do not have plans to add significant functionality to the Python version of the project, as I'm slowly transitioning to using Go for most of everything I do.
Migration from Python gitzebo will be fairly simple when the Go version is available, as Go programs tend to be easier to deploy.
Where Does gitzebo Run?
gitzebo theoretically runs on any host with git and Python installed.
Realistically, it has only been tested on CentOS 6.5. This means that Red Hat Enterprise Linux 6.5 will likely work with no modifications, and that some slight variances in deployment instructions will be needed for Debian-based hosts (including Ubuntu).
I'm working on setting up automated testing across multiple platforms.
The Python Package Index (PyPI) hosts source distributions. The gitzebo
page is here. This means that
easy_install, and the like can be used to install gitzebo.
Documentation will eventually appear on my site.
Prior to installing, you'll need to have
virtualenv installed. On RHEL/CentOS, you can install them with:
# as root user # python-virtualenv lives in EPEL, install EPEL: rpm -Uvh http://dl.fedoraproject.org/pub/epel/6/x86_64/epel-release-6-8.noarch.rpm yum -y install git python python-virtualenv
We also assume the presence of a git user (to run the webapp and to own
the repositories managed by it), and that an
/opt/git directory exists
owned by this user. Make it so:
# as root user getent passwd git || useradd git if [[ ! -d "/opt/git" ]] then mkdir -p /opt/git chown git: /opt/git fi
Before deploying the web app, you need to install gitzebo and initialize
its sqlite database. Note that your virtualenv should be readable by root if
you want Apache deployment to work. Apache needs to read your virtualenv to
start Python interpreters in it -- so we create virtualenvs in
# as git user virtualenv /opt/git/env source /opt/git/env/bin/activate pip install gitzebo gitzebo-schema create
After you've done so, you can bring up a development server to test it out.
This will let you reach gitzebo at
# as git user gitzebo-dev-server
Or you can jump directly to generating a mod_wsgi configuration using a
helper utility. This will let you reach gitzebo at
# as root user # TODO: yet to be implemented rpm -q httpd || yum -y install httpd rpm -q mod_wsgi || yum -y install mod_wsgi gitzebo-generate-conf > /etc/httpd/conf.d/gitzebo.conf service httpd restart
Your Apache instance's configuration directory will vary on Debian or Ubuntu, as will the command to restart Apache.
Once you have an instance running, you can use the username
with the password
admin to log in.
Either way, once you've logged in: change the password.
Why Reinvent the Wheel?!
GitHub, gitorious, GitLab, and gitolite all exist already. Why not use one of these?
The answers come down to requirements:
- easy to set up
- I dislike playing with configuration files for hours at a time to get software working. Applications should work with a minimum amount of documentation reading. Applications should work with a minimum amount of manual steps required to get them running. Ideally, installing an application and deploying it should take one to three commandline invocations.
- easy to use
- If an application is slow, unintuitive, or fails to solve a user's problems, then it will go unused.
- in-house deployment
- I have a bone to pick with software-as-a-service (SaaS). As a rule, I don't like giving up my proprietary datasets. This dislike scales up with business value of the data: I dislike the idea of giving up a folder of dog pictures I've downloaded from the internet to use as wallpapers. You'd think it's useless, and there's no point to being protective: but even that data can be used as a training dataset for artificial intelligences dealing with visual classification... Scale it up to source code, configuration management, and monitoring? Those three are the absolute keys to your IT kingdom. My source code management solution will not be SaaS.
- I don't necessarily want a bug tracker and pretty graphs and the like. The going Linux/UNIX philosophy is to keep each component as lightweight and focused as possible, which is something I like. I took away points for doing too much, which it seemed all git management solutions did.
The four de-facto solutions each violated one of these requirements:
- GitHub is software-as-a-service.
- gitorious is not easy to set up. Deployment on RHEL/CentOS 6 is a pain. You can read more here. I gave up on this approach after a while.
- GitLab was not easy to use due to performance issues. I got GitLab running in a VM with 1GB of memory and a dedicated core. Its performance with two users was slow enough to regularly invoke vulgarities. I don't know if I missed some key setting, but we flagged it a no-go.
- gitolite also failed the ease-of-use test. gitolite does not have a web application for management built in, so I'd have to add one to meet my requirements. As such, it's more of a library or back-end than a full-blown application. It being written in Perl and not having a well-defined API made me extremely nervous, as it seemed like adding a web front-end would be difficult.