We log a lot in our functional tests. In fact, we log so much we wrote our own logger called Herodotus that extends the default Ruby logger so we can have a correlation id added to the log to help keep track of which scenario a given log is from1. But with that much logging, we want to keep things lightweight. That’s where we want to make sure our logs aren’t doing more work than we expect. ...
We’ve written Herodotus1 as a lightweight logger to solve some very specific problems we had in our tests, but we don’t think they are unique to us so we’re going to share with you what we have done and why. Logging to the standard output and the main problem we faced The problem we are trying to solve is one of readability. All good logs are human readable, as when something has gone wrong to the point of a human needing to look in the logs and piece together what has happened we don’t want to make that more difficult than is needed. Now, there are things out there that will help you manage your logs (for example, OpenSearch provides a fantastic toolkit for exploring a large volume of logs), these are often looking at things on a grander scale than we are. There are always going to be times where you want your system to just leave a little message wherever it is currently running, which is why we are looking at things that are logging to the standard output on whatever device they are running. Specifically in our case, we are looking at the output of an automated test pack. ...