When Everything Went Kafka
summary
A practical look at running Apache Kafka: the talk examines Kafka component architecture and collects common operational pitfalls (misleading defaults, client-side surprises, missing docs). It explains the architectural trade-offs behind those issues, discusses why DevOps involvement is necessary, and evaluates whether Kafka’s many configuration options are helpful or harmful.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 1:05Why Kafka in projects
- 3:12Kafka components
- 7:27Cluster architecture and replication
- 10:14Producer internals
- 14:26Consumer internals
- 17:12Controller role
- 18:39Kafka pitfalls
- 20:46Split-brain data loss
- 21:07Replication and leadership issues
- 22:53Default replication settings
- 24:19Broker-client config mismatch
- 25:21Message size errors
- 26:45Batch size vs latency
- 28:31Blocking send issues
- 29:35Consumer driver errors
- 31:41Multi-partition reading
- 34:11Manual disk balancing
- 36:39Topic size limits
- 37:24Zero-copy and compression
- 40:45Finding data in Kafka
- 42:53Timestamp problems
- 44:42Retention and segment errors
- 46:28Consumer offset issues
- 50:20Key takeaways
- 52:13Q&A
related talks
54mWhen Everything Went Kafka: Apache Kafka and the Consumer
16mIntroducing Apache Kafka
46mEmbracing Anarchy
57mDeep Dive into Kafka Architecture: From Simple Scenarios to a Geo-Cluster
46mBrick-by-Brick: Exploring the Elements of Apache Kafka®
51mFive years of highs and lows with Kafka
39m7 Years in Kafka Land
27mKafka - How a Tortoise Helps Us Break Chains
