Here is how we ended up doing it.
- Web Site, seperated out the main home directories, easy enough to do.
- Web Server, this was the tricky one, because it was a compiled executable and when it was updated mostly they were library files. Really we just promoted the new files, even if they were the same as the existing ones, not something I would recommend but it worked at the time.
- Database, sort of tricky but we just used different Oracle instances and that allowed us to test in isolation.
- Other middleware, which was the bulk of the code, we handled this by reinstalling.
To me the whole idea of Risk Management is what needs to be done for every release or build coming into Test, if you don't know your environment enough to see the Risks then you can set yourself up for trouble. Its worth time to ask "What if..?" questions all the time, some Risks you won't know until you have experience to ask the questions, and some will come from doing the releases and seeing the code work in the field. The goal is to minimize those issues that arise from the field, or to eliminate them entirely (a good pipe dream).
Would I do testing like this again? No. I've learned that having your own environment that is yours and yours alone to test with is the best way to mitigate everything. I know it can be done to be in the same environment as Development, I just would never recommend it.
No comments:
Post a Comment