# Graflog

In a distributed system, logging is hard. In order for distributed logging to be
useful, it needs to contain metadata about what caused it. Ideally, a given logged
event should tell us:
- what user input caused it
- what API call caused it
- whether it is an error or not
- its severity
- a correlative id that allows us to find other log events related to it

Graflog is primarily focused on the last piece.