Cluster Governance

Bilibili Tech

Nov 3, 2023 · Big Data

Comprehensive Governance and Optimization Strategies for Large‑Scale Kafka Clusters

To tame a petabyte‑scale Kafka deployment of over 1,000 brokers, the team built a Raft‑based federation controller (Guardian) that adds per‑partition I/O throttling, disk‑aware automatic balancing, multi‑tenant isolation, cross‑IDC migration, request‑queue splitting, tiered storage, auditing, and fully automated rolling upgrades, enabling stable, self‑healing operations.