Our web app source code is stored in Subversion. It's not Java, it's a script app. It needs to be deployed to various servers, and some configuration data needs to be taken into account.
There is no automation at the moment, but I'm about to change that. I'm presently cobbling together a script to build patches for existing installations using the SvnAnt interface to Subversion. Looks like it's got everything I need. http://subclipse.tigris.org/svnant.html I prefer patches because the entirety of the files in the site is simply too large. I think it's not too difficult to come up with a working concept of how state can be determined and steps taken accordingly. State and steps? Well, let's say LIVE has a certain version of the software deployed, which is basically a tag in SVN. In order to bring it up to date, some rather trivial steps need to be performed, basically copy and delete, really not much more in our case. I think my script just needs to know the pattern of the tag for a live version, then find the most recent one (a tag containing a timestamp or something else that is sortable), then determine the files that have been changed with respect to that tag, and also the deleted files, then copy the updated files to the server, delete the deleted files on the server, and perform any build action that might be required. Once I've made sure everything is deployed (maybe some checksumming), I can set the new tag in the repository. Or rather, set the new tag beforehand, and then delete it if the deployment goes wrong for some reason. Any thoughts on such a process you might want to share? Thanks! -- Michael Ludwig
