OpenTelemetry Collector vs Fluent Bit: When to Pick Each
OTel Collector and Fluent Bit get conflated. They are different tools doing different jobs. Picking either-or is usually wrong.
Fluent Bit: log shipping
Fluent Bit: pure log shipper. Tiny memory footprint (~1 MB). Designed for parsing log files, tailing containers, forwarding to backends.
Best when log volume is huge and you need the lightest possible per-node agent.
OTel Collector: unified telemetry
- OTel Collector: receives, processes, exports for metrics + logs + traces. Larger footprint (~50-200 MB) but unifies the three telemetry types.
- Best when you want one pipeline that handles everything OTel.
When to pick each
Pick Fluent Bit alone if: you only care about logs; you need the smallest agent possible.
Pick OTel Collector alone if: you want unified metrics + logs + traces.
Pick both if: high log volume + need for unified pipeline. Common pattern.
Running both well
Standard pattern: Fluent Bit on every node (log collection); OTel Collector centrally (transformation + routing).
Each does what it does best. Memory budget is bounded.
Antipatterns
- OTel Collector on every node for logs only. Wasteful.
- Fluent Bit for traces. Wrong tool.
- Two parallel log paths. Logs duplicated; cost doubled.
What to do this week
Three moves. (1) Trial the candidate tool against one workload for two weeks. (2) Compare against your current using the four criteria above. (3) Plan the migration only if the trial shows real wins, not theoretical ones.