Skip to content

Scaling RabbitMQ#

RabbitMQ is sized by architecture (tier), nodes and storage. Higher tiers and 3/5-node clusters use quorum queues for high availability.

Quorum queues

New queues default to default_queue_type = quorum, so a clustered tier replicates them across its nodes (durable, Raft) without clients passing x-queue-type. Adding nodes increases the number of quorum-queue members; cross-cluster DR is separate (see Disaster recovery).

Tiers#

Tier Storage CPU / memory per node Nodes
Starter 10Gi 1 CPU / 2Gi 1
Small 25Gi 2 CPU / 4Gi 1
Basic 50Gi 2 CPU / 4Gi 3
Medium 100Gi 4 CPU / 8Gi 3
Large 250Gi 8 CPU / 16Gi 3
XLarge 500Gi 16 CPU / 32Gi 3
2XLarge 1000Gi 24 CPU / 64Gi 3
3XLarge 2000Gi 32 CPU / 96Gi 5
4XLarge 4000Gi 48 CPU / 128Gi 5

Clusters use 1, 3 or 5 nodes (odd numbers keep quorum queues available during maintenance).

Change the tier or cluster size#

  1. Open the service and choose Modify.
  2. Select the new architecture and/or node count.
  3. Apply and wait for ready.
nodes = 5
capacity { storage = 250 }
config = { "architecture" = "Large" }

Existing queues and messages are preserved; quorum queues rebalance across the new nodes.

Resize storage#

Storage is per node. Increase it with the tier or set capacity.storage explicitly; persistent messages are retained across the change.

Upgrade the version#

Change kind (for example service/rabbitmq:4.2 → service/rabbitmq:4.3) and apply. The broker performs a rolling upgrade; quorum queues keep the cluster available one node at a time.

Note

Take a backup / verify replication before a major upgrade.