Re: refresh test from prod in wrong way

soalvajavab1_at_yahoo.com wrote:
> in our company everytime they refresh the test from production, all the
> user accounts will be gone!
>
> what is the typical way of refreshing a test db from prod? How do you
> avoid this problem typically?

It really depends.

In an environment with development going on, the test is likely to be ahead of production in terms of table layouts, perhaps some data, and certainly code. A test refresh may involve a layered approach of restoring data from production, then applying changes to the tables and data and upgrading the code.

An environment based on packaged software may be atypical from a DBA standpoint. Oftentimes they have procedures to initially load the data, perhaps using imp, and those procedures pretty much have to be emulated to refresh data, especially if something is being version upgraded.

Production data is not always the best test data. Good test data tests limits, production data may only test limits in the future at the most inopportune time. If you have a lot of data, it may well be worthwhile to create a canonical subset for fast and easy testing. Then you might even be able to give developers their own db to play with, and also have integration and qc databases once they get their stuff working.

I don't know exactly what you mean about user accounts being gone, but perhaps you need to have scripts to add them. I can guess you have one or more schemata with the data in it, and other schemata for the users?  Perhaps you can use exp full=y rows=n to get the users?

Personally, I'm a big fan of using scripts for consistency and repeatability in refreshes. Then again, sometimes I see projects get halted for a year or two, and just finding a script is a lot easier than trying to remember what needs to be done. Happened to me yesterday, in fact, all I have to do is find all the table changes for the last 1.5 years, which is easy because that's just a ls command for a bunch of scripts in a known location.

jg

-- 
@home.com is bogus.
http://www.whatisdeepfried.com/zogg/zogg1.html