Industrial Edge

What problem does it solve?

Manufacturing and OT environments generate sensor telemetry (temperature, vibration, pressure) that must flow from shop floor MQTT brokers into Kafka pipelines, through ML anomaly scoring, and into operations dashboards — often at the edge where latency and connectivity are constrained. Running that stack on Kubernetes at each factory site, while keeping GitOps alignment with a central hub, is the core Industrial Edge problem.

This pattern deploys a repeatable factory stack on east and west spokes: AMQ Broker → Camel K → Kafka → optional KServe → line-dashboard, plus Tekton CI for edge software updates. The hub aggregates metrics (Skupper + Grafana) and exposes unified gateway access without duplicating brokers on the hub.

The Industrial Edge pattern models discrete manufacturing and operational technology (OT) connectivity to Kubernetes: sensors, MQTT, Kafka-centric pipelines, CI/CD for edge software, ML-assisted anomaly detection, and centralized data lakes.

After operators discover Kafka clusters (see AMQ Streams) and mesh namespaces receive ambient labels (Service Mesh), use this page as the business narrative tying workloads together. If you haven’t yet installed, start with Getting Started and scaffold new instances via Scaffolding.

Factory pattern

Factories emit telemetry through MQTT brokers and Camel integrations. Kubernetes namespaces isolate teams while GitOps keeps spoke clusters aligned with approved revisions. Each spoke runs an identical stack independently — the hub aggregates metrics and provides gateway access.

Data stack

Stage	Component	Namespace
Ingress	AMQ Broker (MQTT acceptors), machine sensors	`industrial-edge-tst-all`
Integration	Camel K routes (MQTT→Kafka, Kafka→S3)	`industrial-edge-tst-all`
Streaming	AMQ Streams / Kafka dev-cluster topics	`industrial-edge-tst-all`
Replication	Strimzi MirrorMaker (factory→data-lake)	`industrial-edge-stormshift-messaging`
Data lake	prod-cluster Kafka + Camel S3 archiver	`industrial-edge-data-lake`
Processing	OpenShift AI (KServe InferenceService)	`industrial-edge-tst-all`
CI/CD	Tekton pipelines (buildah + deploy)	`industrial-edge-ci`
Visualization	line-dashboard (WebSocket consumer)	`industrial-edge-tst-all`

Camel K integrations

The mqtt-to-kafka integration bridges sensor data from the AMQ Broker to Kafka topics. It runs as a Camel K Integration CR deployed by the industrial-edge-tst chart. The Camel Kaoto software template in Developer Hub scaffolds additional Camel routes with a DevSpaces-ready project including:

MQTT→Kafka route (temperature, vibration)
Kafka→S3/MinIO archiver (data lake)
Alert→Mailpit route (anomaly notifications)
Kaoto visual editor + Continue AI code assistant

To scaffold a Camel route: Developer Hub → Create → Industrial Edge — Camel Routes (Kaoto + Continue AI) → choose east or west.

Spoke components (per cluster)

Each spoke (east/west) deploys the following Argo CD Applications:

Application	Chart	Purpose
`industrial-edge-tst`	`charts/all/industrial-edge-tst`	Sensors, broker, Kafka, Camel, dashboard, ML
`industrial-edge-stormshift`	`charts/all/industrial-edge-stormshift`	Factory Kafka + MirrorMaker
`industrial-edge-data-lake`	`charts/all/industrial-edge-data-lake`	prod-cluster + S3 archiver
`industrial-edge-pipelines`	`charts/all/industrial-edge-pipelines`	Tekton build-and-test pipeline

Service mesh (ambient mode)

Industrial Edge namespaces run under OSSM3 ambient mesh with ztunnel L4 metrics. The namespace industrial-edge-tst-all carries the label istio.io/dataplane-mode: ambient. Ztunnel captures istio_tcp_connections_opened_total and related L4 series; L7 istio_requests_total flows through waypoint gateways where deployed.

Exception: spoke-gateway-system and industrial-edge-data-lake stay off ambient to avoid TLS conflicts with MinIO and WebSocket routing.