Tag-based logging

The discussion of query semantics woven throughout makes this very real, by my reading. This has given me much to think about. The remark about Grafana and friends being a surrender is hard to admit to.

I’ve been happy using tracing (opentelemetry) and shipping them to CloudWatch. It’s amazing how nice it is to debug things when you have the whole request execution tree available instead of a linear list of log lines.

(This fits request-driven services best.)

I’m waiting for a future day when the traces get too heavy or expensive. It hasn’t arrived yet. 🙂

Even advanced logging systems that can tune levels per component can’t efficiently select only performance-related messages.

That seems fixable :/ https://github.com/backtrace-labs/dynamic_flag can toggle branches based on a tag (only one) and a regex match on source location. The main motivation for that feature was granularly enabling detailed logs in stateful services that could take a few hours to boot up. The self-modifying code is just a bonus, you can achieve the same with static registration of granular enablement state (could be as small as one bit per log statement).