Frequently Asked Questions

There currently aren't many, but here's a start:

1. Is AutoPatch tested well?

   We develop AutoPatch using JUnit and Clover, a coverage utility. Our unit tests contain a *lot* of assertions so we know the tests themselves are useful, and you can see the coverage for yourself:

   

   In short, we believe AutoPatch is tested well (especially the complicated bits), and is very robust. Our experience with it in production bears that out.

2. Is anyone using AutoPatch?

   AutoPatch is in production use at almost every one of our clients. Something important to note is that the total headcount of DBAs at all the clients we have that use AutoPatch is zero. Developer throughput is high though. That's efficiency.

3. What do DBAs do if they don't move patches around?

   We believe DBAs should be involved in patch creation (where knowledge of SQL and SQL programming languages like PL/SQL is very important) and in database tuning. Synchronizing environments and applying patches is something computers should be good at, DBAs shouldn't have to to be involved.

4. How do you manage patches on branches when you have emergency production changes?

   This is a tough one but can be handled well. The best way we've heard so far is to implement two separate patch streams (different system names).
   You should configure the patch streams so that you can run them in distributed mode or with both as standalone.
   One patch stream is used for emergency patches on branches only.
   Any time you cut a branch, leave a gap in the patch sequence on the main patch stream.
   Any time you make an emergency patch, put it in the emergency patch stream on branch and trunk using the lowest available number from the gap.
   Any time you make a normal patch on trunk, put it in the main patch sequence using the first available patch number after the gap.
   During normal development, execute the two patch streams individually as standalone patch streams
   During full database rebuilds, execute the patch streams as a distributed system.
   This guarantees that emergency patches propagate to all systems immediately, but also guarantees that database rebuilds apply patches in the same order they applied when they were created.
   In the future we'll probably support a dot-notation (similar to CVS revision branch numbering) so you don't have to worry about leaving a patch sequence number gap.