Do not try to reinvent the wheel, so I wonder if any registration system supports what I propose to do.
Background:
I am working on an extremely large system where tens of thousands of users access servers at any given time. There are many surrounding infrastructures, so you can imagine how researching rare errors looks like by simply reading the magazines in such an ecosystem.
Our system uses log4j.
The main root of the problem was that the original developers were quite “economical,” to put it mildly, when confronted with an unknown error. Of course, the DEBUG level would be useful in most cases, but, of course, production logs are set only at the ERROR level: (
Now I am trying to create a better world for our children and I want to expand the journal system to be more “forgiving” regarding the level of the magazine.
What I mean:
Since the logs of the DEBUG level logs (N before and M after the ERROR log) can contain important data, why not lower the log level in the system around ERROR.
My idea:
- Assume that the ERROR log level is set for the system.
- use some sort of rollover buffer size X. Each log is sent there and forwarded to log4j.
- ERROR, N M
, .
?