I recently had an interaction with an NServiceBus customer and noticed that they are pushing the log of their import process all the way to the UI. While this may be normal in some cases, but why do you put all the technical details, including exceptions, stack trace, in case of errors in that customer facing log window? The answer was: it is hard for us to gather all the relevant information from a distributed process that’s running in a handful of microservices. When something goes wrong, it is hard for us to gather all the relevant information.
It all boiled down to this: while each individual service capture all the detail of a successfull and failed job, the logs go into individual log files on those services. It is hard to connect these logs across different services. In a nutshell, distributed logging is hard. But does it have to be? Continue reading