An update on Khepri adoption in RabbitMQ
summary
Khepri is a Raft-backed metadata store that already replaces most Ninja-backed stores, drop transient entities (no transient queues/exchanges, exclusive queues remain), and moves topic bindings, imports and recovery into a single Raft cluster for much faster recovery — benchmarks show up to a 70% speedup on some exchange operations and large imports (1,000 queues × 100 topic bindings). Khepri adds a local-cache mode to match Nnesia’s ETS lookups, changes snapshotting to reduce memory spikes, and will require rolling Erlang 25; adoption is via a feature flag in 3.x or a blue‑green migration to 4.0, and cluster operation now strictly needs a majority.
AI-written summary · about
chapters (5)
related talks
40mKhepri - Mnesia replacement
40mPerformance Improvements
38mRe-architecting Apache Kafka for the Cloud
35mMigrating from Kafka to RabbitMQ at SimpleBet: Why and How
40mOpen Riak how we gained from leaning on the BEAM
31mRiak's Rebirth
01h 00mRabbit MQ vs Kafka: When, Which, Where?
39mKafka-Backed Elixir at Scale