Files project dev server
Follow-up to Getting a project up and running from Kieran's blog
One of the important things to try and get ready as early as possible is a test/pre-production system that is fairly close to what you expect your live environment to be. This is so that you don’t spend the whole time developing on a single JBoss instance on a single processor Windows box with local storage and then deploy on a multi-processor, multi-JBoss and remote storage box and discover that nothing works!
We are now starting to run Solaris 10 which gives us the great Zones feature. Our sys admins have setup a zone on one of our new boxes that is a test/pre-prod environment for the files project. We’ll run something like this:
- Single Apache 1.3x instance with HAProxy to load balance between the JBoss instances
- Two JBoss instances both running live rather than a live and a standby
- For now local storage, but eventually we’ll use our NetApp
The twin live JBoss instances means that our application will have to be completely stateless. This is a good thing for scalability, but it makes multi-step processes within the application a bit harder as we won’t have a session to store data in. This is usually not a problem for simple applications, but working on something like a mutli-step zip upload could be tricky.
The other advantage of having a test instance up and running is that you can start to point very early test users at it (rather than a local instance on your own machine). This gets you some good early feedback/ideas/bug-spotting.
Add a comment
You are not allowed to comment on this entry as it has restricted commenting permissions.