search

cncf / tag-observability

Technical Advisory Group for Observability 🔭⚙️
https://cncf.io/projects
Apache License 2.0
640 stars 83 forks source link

whitepaper: Add clear example on why we have more than 1 observability signal. #119

Open bwplotka opened 1 year ago

bwplotka commented 1 year ago

Epic idea from @beorn7 to dive deeper on why we have so many signals:

Björn Rabenstein: we can't have a single bicycle that works efficiently on both asphalt roads and off-roads-- How about explaining this as logging being the "first principle signal". You could model all the other signals as logs (examples: log all the spans with their context and let log processing figure out the distributed traces — instead of collecting metrics, dump their values into the log file every minute and let the log processing figure out the rest). Metrics and Traces are essentially optimizations for a specific shape of data and data usage. This optimizations is necessary because logging all the things is just waaaay to expensive.