Fitting Stateful Functions into a streaming platform
summary
Nordstrom runs a Kafka-based streaming platform handling over 5,000 distinct resource changes per day and uses Prometheus for telemetry. Flink Stateful Functions become the event-driven control plane that stores snapshots to compacted topics, enforces credential expiry, and executes topic self-delete policies (TTL and log-size or rate thresholds). Demo shows credentials auto-expire, topics auto-delete, and snapshots upserted back to the metadata store, replacing many scheduled jobs and simplifying API services.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:33Nordstrom
- 1:06Why a streaming platform
- 3:19Platform interactions
- 4:11Application applications
- 4:41Telemetry
- 6:59Stateful Functions
- 8:07Blank Stateful Functions
- 9:34Kafkatopic Function
- 10:54Snapshots
- 11:38Metrics
- 13:37Stateful Match
- 15:40Demo
- 24:04Deleting topics
- 24:41Additional platform behaviors
- 27:01Outro
