Rainy day… Thoughts arn’t well aligned… At the end of the day I was wondering on the Internet searching for an inspiration. For a moment I thought I found something interesting. It was on the Jonathan Lewis’s blog - something about Oracle traces. But just a second later after finished reading I grasped what the message was about. In few words there was confusion, doubt and something else, or so it seemed. The title was Trace Files. And here you go an inspiration came, and as usual from a direction you’d never expect.
Should we use undocumented or unofficial methods? Yes, no doubt. That is of course when they are judicially legitimate. I just wanted to point out few key elements of the documented versus undocumented approach as I understand and interpret it. In a different scope this might become discussion about supported versus unsupported and so on.
Documentation is a right place to go if it exists. If it doesn’t, “hacking” in the right meaning of the word, is the way to know the system better than it can possibly be documented. Consider Oracle tracing facility as free of charge and supported debugger. By using it you’d learn about the Oracle and application that uses Oracle, about specifics of different Oracle versions, the improvements in newer releases and deficiencies in former releases.
What you would lack when you use Oracle tracing is official evidence. If you see a problem in the trace, but struggle to reproduce it using documented forms of proof, you might have some trouble building a case for an SR. You also will have certain difficulties explaining it to the Oracle TS what is that you are trying to solve. And final point against it, when filing SR, is that in Oracle R&D work people just like you. Reverse enginiring an issue from a trace file that you attached is the last thing they wanted to do. Similar to any other trace whether that is a stack trace from JVM or a core dump. It is always better for them to have a reproducible at will test case. However, many times by looking at traces I found things to build a simple test case that I wouldn’t have gotten that easily otherwise.
To summarize, as far as I’m concerned, if the Oracle traces, considering their generation can be done using supported API, allow me to find true problems and do right decisions I’ll keep using them even if their format is neither documented nor supported.