Logging Practices Debate

Sometimes logs are the only way.It's old but reliable.

A concrete example of why its not always terrible to have logs data sitting around.

> wait a minute, you're going to blow up our logs. You really need to have that pointed out to you?

Parsing logs isn't the only way to troubleshoot.

Isn't logging enough to track down a lot of such problems?

It's time to stop logging at scale. Logs are just not worth it. Setup telemetry instead, such as statsd. Select actionable metrics, such as the count of successful and erroneous operations. Save disk, save I/O performance, never stress about log leaks again.

Think this mostly boils down to the opinion that you should handle errors properly or ignore them instead of just logging them. This article might read better if you defined what "logging" means up front and make sure it's a well scoped definition. I can also think of a bunch of cases where logging is super important that aren't really covered by any of your arguments.For example, if you have a micro services architecture in production in which many services might be acces

Don't forget the logs produced by the logging infrastructure.

Lots of people do exactly this:https://tersesystems.com/blog/2019/10/05/diagnostic-logging-...

I think the author raises some valuable points but does not emphasize the key point about logging; that _logging has relatively limited applicability_. If you think of logging that way, you log everything, but keep only a time-bound window of log data. Around two weeks is plenty because most of the times you need to access the detailed log information is same or next day. I think careful attention should be paid to annotating log messages in various parts of the stack with some piece of data tha