Observability with Spring-Based Distributed Systems

[kymr]

Ref

• https://www.youtube.com/watch?v=PM0mO4IxpSE

Observability

• Log
  • easy to "grep", manually read
  • time and write formatted or structured logs
  • ubiquitous apis, but requires coordination
  • pull raw events into a parsing pipeline
  • don't log irrelevant data, filtering
• Metric
  • can identify trends
  • time and store the number
  • easy, but least context
  • report duration buckets near-real time
  • read-your-writes, coarser grain
• Trace
  • identify cause across services
  • start, propagate and finish a "span"
  • hardest, as identifiers must be passed within and between services
  • report spans near-real time
  • sampling, but needs to be consistent

How to not see tracing

• Buddy
  • another process intercepts yours
• Agent
  • code patches code
• Framework
  • code intercepts or configures code