In this episode, you'll explore with
@HenrikRexed the exciting new features of Fluent Bit v4, announced at KubeCon Europe 2025. Fluent Bit has long been a powerhouse for log processing, but with v4, it’s stepping into the world of distributed traces and metrics like never before.
🔍 What You’ll Learn:
- All the major updates in Fluent Bit v4
- How to use the new conditional processors for smarter log pipelines
- Deep dive into trace sampling processors: probabilistic vs. tail sampling
- Real-world benchmarking results comparing Fluent Bit and the OpenTelemetry Collector
- Limitations and missing features you should know before going all-in
💡 Whether you're building observability pipelines, optimizing resource usage, or just curious about Fluent Bit’s evolution—this episode is packed with insights, demos, and practical advice.
📊 Benchmark Highlights:
- Fluent Bit’s tail sampling can consume up to 1GB of memory—why this might be a red flag
- CPU usage increases by 250% with sampling—but still within acceptable limits
- OpenTelemetry Collector outperforms Fluent Bit in sampling stability and telemetry
⚠️ Key Takeaways:
Fluent Bit v4 is a big leap forward, but still lacks:
- Kubernetes metadata enrichment for traces/metrics
- Metric conversion capabilities
- Trace-aware load balancing for horizontal scaling
- Telemetry visibility into sampling decisions
---------------------------
📖 Chapters 📖
00:00 Intro
00:13 Welcome
02:20 What’s New
06:49 Conditional processor
11:24 Sampling processor
20:38 Benchmark
24:10 Conclusion
-----------------------------
🔗 Additional Links
Dynatrace Trial
Github repo for the tutorial
Fluentbit
Blog
-----------------------------
Subscribe to our YT channel
Stay up-to-date with Dynatrace! Follow us on Facebook, Instagram, LinkedIn, Twitter, Twitch
