Strong article, but you missed a key point on quantity and quality!

The trouble with logging is most programs have too much logging that is also not useful enough. I worked for a company whose flagship program generated 2 GB of logs a day, and this was 10 years ago. Yet nearly all of those logs were garbage and there was no way to parse those logs to extract useful information, as they were written in freeform text.

The one use they had was when something got out of whack, then some poor sucker had to wander through millions of lines of logs, trying to see if anything useful were in there. But they reacted with great hostility to any idea of of compressing, standardizing or slimming down the log.

In a well–running system, people periodically look at the logs, remove unnecessary logging, and add new logging. But most of the time, this is just sloughed off as technical debt.

Thanks for an excellent article!