Overview
Boomi processes moved regulated transactions across systems, with RabbitMQ as the transport between them. Recovering from a data center event meant a manual promotion nobody had practiced, and once a message was consumed there was no way to replay it.
Challenge
Two gaps compounded each other. Failover depended on a human following an unrehearsed procedure, and because queues discard on consume there was no audit-friendly way to reconstruct what had flowed during an incident. Both had to be solved without breaking the controls the estate is audited against.
Environment
Boomi integration processes over RabbitMQ on a container platform, spanning a primary and a disaster recovery data center.
Approach
AceMQ produced a disaster recovery design covering manual and automated failover, including what the commercial RabbitMQ distribution changes about the automated option. In parallel we assessed converting selected queues to streams so messages remain replayable after consumption, and specified a buffering layer so Boomi processes degrade rather than fail when the broker is unreachable.
Solution
- Disaster recovery design comparing manual promotion against automated failover
- Cluster sizing across the primary and DR data centers
- Queue-to-stream conversion assessment so messages stay replayable after consumption
- Review of replay against the estate's audit and retention controls
- Buffering layer so integration processes degrade instead of failing during an outage
- Operational runbook and training for the support and development teams
Outcome
The integration estate gained a rehearsed recovery procedure, a defensible answer to what happened during an outage, and a runbook its own engineers can execute without waiting on a vendor.
Technologies
Boomi
RabbitMQ
Kubernetes
IBM MQ
PostgreSQL
