Durable MQTT

What must survive a broker restart?

Recover broker and delivery state through PostgreSQL when a restart must preserve your application’s messages.

Teams responsible for recovering device messages after a restart.

Who is Pro for?

Teams responsible for device message recovery need a clear answer when a process exits, a host restarts or a client reconnects. Pro provides a durable single-active-node MQTT profile and standard installation/configuration/defect support for teams operating their own infrastructure.

What survives a restart?

With the configured durable acknowledgement policy, PostgreSQL stores broker sessions, retained/offline state and the delivery frontiers needed for recovery. The optional bridge outbox records pending external delivery work. Checkpoints and replay must use the matching store, configuration and image.

You supply PostgreSQL, verified trust/credentials, backups and the required durability settings. Test the exact workload through an isolated restart and confirm application and sink receipts.

The durable recovery skill explains the boundary and the evidence to retain.

Which clients and destinations fit?

Pro adds secure WebSocket to TCP/TLS. It includes file, verified mTLS and JWT/JWKS authentication surfaces, topic policy, rate controls, retained messages, Wills and sessions.

Current-source MQTT 3.1.1 supports QoS 0/1/2; MQTT 5 uses the declared QoS 0/1 profile. Check MQTT coverage.

Route selected topics to baseline PostgreSQL, HTTPS webhook, MQTT or your own C adapter. Bridge-only routing can use the durable outbox. See Build for package wiring and commitment/retry requirements.

What does the service include?

The subscription service unit is a production node, with development/staging for the same workload, published release/security updates while active, and standard 8×5 installation, configuration and defect triage. The initial-response target is two business days. An uptime or time-to-fix SLA is not included. Review the agreement and checkout.

Production use is permitted under the license grant. Device, connection and message counts are not pricing meters.

When should I select another tier?

Choose Max for transformation/schema workflows, industrial acquisition, device gateways or the provider adapter catalog. Choose Enterprise for fenced standby recovery and governed operator changes. Pro’s bundled panel and API expose operational reads by default; the selected configuration blocks governed writes.

Release-candidate scope and published evidence remain explicit. The measured Max export workload does not establish Pro’s durable capacity. Size and verify the exact installed profile before assigning production traffic.