DevTeach: Home Grown Production System Monitoring and Reports
That was an interesting talk, Owen Rogers talked about how to setup an Operation Database in order to get more visibility on your production system. This is the first time that I sat in a talk that is a "Release It!" influenced talk, and it was very interesting.
The type of Operation Database is focused on adding more information for the developers, rather than exposing more information to the operation team. What was especially interesting is that the amount of data being capture is very small. The standard log data (time, message, exception, etc) and a command log, which I think about as a message handling log. This log capture some statistics about messages / commands that the application has handled. Message size (in & out), time, processing time, etc.
From that, you can get pretty interesting data about your application (just showing avg. message processing time over a period of time is extremely valuable).
The nice part about this is that the entry cost is basically zero.
Comments
This is a very interesting topic. I had some similar thoughts about gathering information about the application that I am currently working on. To get information about non-functional numbers as the user is acctually running the software. This should help us developers address issues with numbers backing us up that are not just profiling on dev machines, but acctual information of the system at work.
Do you know if there is a transcript of the talk that you were at?
Shouldn't this sort of information be captured on a machine that is NOT production?
No transcript that I am aware of, but you should be able to get the presentation from dev teach
That is the idea, it is captured on production, and available outside it
Comment preview