I had forgotten that I wrote this 10 years ago while I worked at Bitly. Looks like the old Bitly tech blog isn't much maintained, but there's still some gems there:
Getting more clues from Python’s logging module
Not a bad little article. Nowadays we tend to rely more on third-party tools like Datadog which sometimes offer similar functionality out of the box, but sometimes you might end up rolling your own.
Other fun Bitly posts
While there I ran into some other old goodies:
Adventures in Optimizing Text Processing - by my old tech lead Michael Richman. I remember working on some of those Hadoop jobs!
dablooms - a cool approach to solving spam at scale.
Metrics - Building Clickatron and sortdb - static key value database - I can tell you that clickatron was a bit weird to work with as a developer writing queries against it, but it's really impressive how Bitly was able to build a metrics system that scaled to 100 billion data points (as of 2012) with such a small dev and infrastructure team. There really was nothing suitable off the shelf at the time, so they had to invent it. Some of the choices made were things I found really shocking at the time... what do you mean, you built a timeseries database that large out of plain text files? And you rebuild the entire database from scratch every night? Wat? But it worked well!