The author argues that monitoring and observability tools are not as simple as they appear, as they require constant maintenance and may not scale. Logs, while a good idea in theory, often end up being useless and expensive to maintain. Metrics can be a better option, but they also pose scalability and complexity challenges as they become more useful outside of just debugging applications. The author suggests setting clear retention limits, considering SaaS options, and defining an acceptable end state to avoid costly and overwhelming observability systems.
https://matduggan.com/were-all-doing-metrics-wrong/