A simple Phing-based deployment script. Written in a few hours, I wanted the script to adhere to a few key principles:
- Atomic - no partial uploads
- Git-based - deploy the latest tagged release, downloaded from a repo
- Tidy - cleans up after itself (both on local and remote)
After calling phing, the following things happen:
- The repo is downloaded into a temporary directory using
git archive - An "install" script is created in the temporary directory
- A temporary directory is created on the remote host
- The install script and archive are pushed up to the remote host
- The install script is executed remotely
- The install script unpacks the archive and rsyncs the temp directory's contents into the live project folder
- The "current" symlink is deleted and repointed at the latest release
- The install script is deleted as are all temporary directories using a safe
rmdircall
The script must be called from within its directory. By that, I mean you can't currently call it like:
phing -f my/project/directory/build.xml
You must be in my/project/directory and call phing
As this script currently deploys the entire repo, there's no scope for pushing up local config files. I wouldn't commit sensitive information (passwords or keys) to the repo, so those files must be sourced locally, rather than from the repo. This is a very simple thing to fix, I just need a few moments and some real use cases to test it out on.