Deploy and scale
Running OPC UA for two hundred machines rather than two. Containers without breaking certificate trust on every restart, redundancy and aggregation, admission control, connect storms, and the subscription arithmetic that decides whether a plant rollout works.
| Lessons | 10 |
|---|---|
| Time | about 80 minutes |
| Prerequisites | A4; B3 helps |
| Cost | Free: written lessons open, video and labs need a free account |
What you will be able to do
- You can containerise an OPC UA client without regenerating its identity on restart.
- You can design for a plant, with the subscription arithmetic written down.
- You can avoid a connect storm after a network event.
- You can decide between direct connections, an aggregation server and a gateway.
Syllabus
-
1
Containerise without losing identity9 min
The trap that catches everyone: a non-persistent pki volume regenerates the application certificate on every restart, so every peer's trust list goes stale on every deploy.
Asset A container losing trust on restart, then the volume mount that fixes it, with both logs
-
2
Configure for a container8 min
Hostnames, ApplicationUri and the SAN inside a container network, where nothing resolves the way it does on a laptop.
Asset A working Dockerfile and compose file for a client, with the certificate configuration
-
3
Do the subscription arithmetic9 min
Machines times tags times rate, against MaxMonitoredItemsPerSubscription and the server's session limits. Done on paper before the rollout, not during it.
Asset A worked capacity model for 200 machines and 12,000 tags, with the server limits that bind
-
4
Stagger your connects7 min
The most effective single client-side measure against a connect storm after a network event.
Asset A 200-client reconnect with and without stagger, with both server load traces
-
5
Choose direct, aggregated or gateway8 min
Three topologies, with what each costs in latency, failure modes and licences.
Asset The same plant drawn three ways, with the failure mode of each marked
-
6
Run redundancy honestly8 min
Server redundancy, client redundancy, and what actually fails over, which is less than the marketing suggests.
Asset A failover test with the data gap measured
-
7
Apply admission control7 min
Protecting a server from its clients: session limits, monitored-item limits, and what the server does when it says no.
Asset A server hitting each of its limits, with the status code returned for each
-
8
Monitor the collector8 min
What to alarm on so a stopped data flow gets noticed in minutes rather than at month end.
Asset A monitoring checklist with the six signals that catch a silent collector
-
9
Deploy to the edge8 min
Constrained hardware, intermittent links, store-and-forward, and what OPC UA does not buffer for you.
Asset An edge deployment with a simulated 20-minute outage, showing what was and was not recovered
-
10
Upgrade without an outage8 min
Firmware and server upgrades that shift namespace indexes or change the model, and the GeneralModelChangeEvent that warns you.
Asset A pre- and post-upgrade nodeset diff, with the client changes it required
Lab
Free account required
Run a client in a container against a local server, restart the container, and prove the trust relationship survived. Then break it by removing the volume and observe the exact failure.
Unlock the lab →Last verified 7 September 2026 against OPC UA 1.05.07. OPC UA Connect is an educational resource and is not affiliated with the OPC Foundation.