Deploy Ignition for Cluster
This article refers to Platform v3.4.0. The current Platform version is v3.4.0.
Overview
A single Ignition or Ignition Edge gateway represents a single point of failure. If the gateway becomes unavailable (whether due to maintenance, hardware failure, or an operating system crash), client sessions disconnect and data collection stops until service is restored.
Ignition for Cluster and Ignition Edge for Cluster provide high availability through a Hot Standby redundancy architecture. Two Ignition gateways run side by side: a Master gateway that serves all client traffic and a Backup gateway that continuously synchronizes its state with the Master through the Ignition Gateway Network.
A Traefik reverse proxy sits in front of both gateways and exposes a single endpoint to users and applications. If the Master becomes unavailable, Traefik automatically redirects traffic to the Backup while Ignition promotes it to active status. As a result, clients experience only a brief reconnection and no manual intervention is required.
Barbara automates the deployment of the entire stack, including the Master gateway, Backup gateway, Traefik, and the initialization service required to prepare the cluster. Once deployed, Ignition's redundancy engine continuously synchronizes both gateways and manages role transitions automatically.
How high availability and failover work
Gateway failover
Traefik monitors the active gateway through Ignition's /StatusPing endpoint every 5 seconds, with a 3-second timeout. If the health check fails, Traefik automatically redirects HTTP traffic to the Backup in approximately 5–8 seconds.
Ignition's redundancy engine then completes the failover process by promoting the Backup to Master. This promotion typically takes an additional 30 seconds, ensuring the newly promoted gateway assumes control with a fully synchronized state.
Because traffic routing and gateway promotion are handled independently, clients typically experience only a brief reconnect while the cluster continues operating normally.
TCP protocol failover
Industrial protocols such as OPC-UA, MQTT, Modbus, BACnet, and DNP3 are exposed through Traefik and distributed between the Master and Backup gateways while both are healthy.
If one gateway fails its connection-level health check, all new protocol connections are automatically routed to the remaining healthy gateway. Existing sessions may need to reconnect, depending on the protocol and client implementation.
Infrastructure requirements
The Ignition redundancy pair provides application-level high availability: gateway services remain available when a node or gateway fails. However, the redundancy pair alone does not guarantee full cluster-level high availability.
Docker Swarm relies on the Raft consensus protocol, which requires a majority of manager nodes to remain reachable. In a two-manager cluster, the required quorum is 2 out of 2. If either manager fails, the remaining node continues running existing workloads, but the Swarm control plane can no longer schedule, update, or recover services automatically.
For this reason, production deployments should use three Swarm manager nodes. With three managers, quorum becomes 2 out of 3, allowing the cluster to tolerate the loss of any single node while remaining fully operational.
Our cluster applications therefore run best on a Barbara cluster with three active nodes. If your cluster does not exist yet, follow the Clusters guide to create one and ensure all nodes are in Active availability status.

A three-node cluster ready for deployment
In this tutorial, you will deploy Ignition for Cluster, verify that redundancy is operating correctly from the Ignition Gateway UI, and configure both gateways from the application settings instead of the Designer. The same steps apply to both Ignition for Cluster and Ignition Edge for Cluster.
Add Ignition for Cluster to your Library
If the edition you want is not yet in your Library, add it from Barbara Marketplace:

Ignition for Cluster
Ignition for Cluster packages the full Ignition SCADA platform as a high-availability deployment, running two redundant instances in your cluster for continuous, uninterrupted operation.

Ignition Edge for Cluster
The lightweight Ignition Edge edition deployed as a Hot Standby redundancy pair, bringing high-availability local SCADA and IIoT to edge sites.
For the full add-to-Library flow, see Add a Marketplace app to your Library.
Create cluster volumes
Before deploying Ignition for Cluster, create one persistent volume for each service that keeps state: Traefik, the Master gateway, the Backup gateway, and the configuration service that applies your Application Config settings. Open your cluster from the Clusters list and go to Cluster Info. In the Volumes card, click + to add a new volume.

The Volumes card in Cluster Info
Create the following four volumes. In the Add new Volume dialog, enter each name and click Add:

Add new Volume dialog
| Volume name | Used by |
|---|---|
Ignition_vol_01 | Traefik |
Ignition_vol_02 | Master gateway (gateway) |
Ignition_vol_03 | Backup gateway (gatewayr) |
Ignition_vol_04 | Configuration service (ignition_config) |
Once created, each volume appears in the Volumes card with status Not in use, which is expected before deployment:

Cluster volumes created
Open the cluster and start the wizard
Open your cluster from the Clusters list in the Barbara Panel.

Cluster detail view
Click Deploy to open the installation wizard.
1) Command settings
Select the application you want to deploy (Ignition for Cluster or Ignition Edge for Cluster) and the version, then click Next. The remaining steps are identical for both editions.

Selecting Ignition for Cluster
2) App secrets
The App secrets step has one tab per service: traefik, gateway, gatewayr, bbrinit, and ignition_config. For this tutorial you change three settings; keep the defaults for everything else.
On the traefik tab, set TLS_MODE to selfsigned. Traefik generates a self-signed certificate and automatically redirects HTTP traffic to HTTPS. This is sufficient for clusters running on private networks, where no public domain or external certificate authority is required.

App secrets: traefik tab
On the gateway tab (the Master), set ACCEPT_IGNITION_EULA to Y. The Master will not start without it.

App secrets: gateway tab
Leave the gatewayr (Backup) tab unchanged.

App secrets: gatewayr tab
There is nothing you have to set on the bbrinit tab.
Even so, the Application Config secret field is worth knowing about. Barbara turns it into the access key the configuration service uses to talk to the gateways, and generates that key for you when the field is empty. Set it to a phrase of your own, at least 24 characters, when you want the key to be reproducible: derived from a phrase you keep, it can be regenerated, which is the way back if the configuration volume is ever lost while the gateway volumes survive.
The rest of the tab is optional too. To start the pair from an existing gateway backup, set SOURCE_URL and enable BOOTSTRAP, following the same process described in Deploy Ignition. To give the gateways credentials they need without writing them into the configuration, such as the password of an OPC UA server it connects to, set Gateway secrets as described under Secrets.

App secrets: bbrinit tab
Leave the ignition_config tab unchanged. Its only setting, IGNITION_API_TOKEN, is for the case where you bring your own gateway backup that already carries an access key and want to keep that one.

App secrets: ignition_config tab
Every configuration variable and its default value is documented in the application's technical notes on the Barbara Marketplace.
3) Compose config
In the Compose config step, assign each service to the volume you created for it. Set VOLUME_TRAEFIK to Ignition_vol_01 on the traefik tab. Keep the port mappings as they are.

Compose config: traefik tab
Set VOLUME_IGNITION_GW_DATA to Ignition_vol_02 on the gateway tab.

Compose config: gateway tab
Then set VOLUME_IGNITION_GWR_DATA to Ignition_vol_03 on the gatewayr tab.

Compose config: gatewayr tab
Finally, set VOLUME_IGNITION_CONFIG_STATE to Ignition_vol_04 on the ignition_config tab. This volume records which settings Barbara already applied to the pair, so the service does not start from scratch if the cluster reschedules it onto another node.

Compose config: ignition_config tab
The Backup is also reachable directly, bypassing Traefik, on the port set by Backup gateway direct port in the previous step. It defaults to 8089 and is useful for diagnostics when you need to reach the Backup without going through the proxy.
4) Deployment & Replicas
There is nothing to change on this step. Leave gateway, gatewayr and bbrinit in the default Replicated mode with 1 replica each.
The traefik tab also shows Replicated, and you can leave it that way: the application forces one Traefik instance per node regardless of what this step says, through a reserved setting that accepts no other value. That is what keeps the entry point answering wherever the virtual IP currently is, even if a node goes down.

Deployment & Replicas: gateway, gatewayr, and bbrinit
5) App config
The editor opens already filled with the Ignition template: the sections you can declare, each one empty. This is where you configure the gateway pair without opening the Designer, and you can come back to it at any time after the deployment.
Leave the template as it is for now and click Send App. Barbara dispatches the deployment to the cluster, and the Application Config section below covers what to put in each part once the pair is running.

The App config step
Wait for both gateways to start
The deployment now runs automatically. The initialization service prepares the environment, and both gateways start, initialize, and synchronize with each other.
Give it time. The primary gateway is usually serving within 3–5 minutes, but the backup waits for the initialization service to finish preparing its own volume before it starts at all, and on a cluster with shared storage that wait can run considerably longer. The workload card reports no error while this happens, because nothing has failed: the backup simply has not started yet. You can monitor the deployment from the Workloads tab; the workload card shows a Services section with the status of each service.

Workload card with all services running
Access the Ignition Gateway UI
Open the Ignition Gateway UI from the URLs section of the workload card, which provides HTTP and HTTPS access through Traefik, together with a direct link to the Backup and the exposed protocol endpoints:

The workload URLs panel
- Ignition Gateway UI (HTTP): routes through Traefik to whichever gateway is currently active.
- Ignition Gateway UI (HTTPS): HTTPS entrypoint (active when
TLS_MODEis notoff). - Ignition Backup Gateway: direct access to the Backup on the port you set as Backup gateway direct port (
8089by default), bypassing Traefik. Use this for diagnostics when you need to reach the Backup specifically.
Log in with the credentials set in App Secrets; the defaults are bbruser / bbrpassword.
Verify Ignition redundancy
With both gateways running and communicating over the Gateway Network, Ignition automatically elects a Master and keeps the Backup synchronized. You can verify the redundancy state from the Gateway UI.
In the left navigation, go to Platform → Redundancy.
The Master shows Gateway Role: Master, Activity Level: Active, and Sync Status: Good with the peer connected:

Redundancy on the Master
Open the Ignition Backup Gateway URL to verify the Backup. On first startup it may show Sync Status: Out of Date. Once synchronization completes, the status changes to Good:

Redundancy on the Backup
Configure both gateways from Application Config
The pair is running. Every device connection, every link to another gateway and every tag provider would still have to be entered by hand in the Designer on the Master, and entered again the next time you rebuild the cluster.
Application Config removes that step. You declare what the gateways should have in a JSON document in the Barbara Panel, and the configuration service keeps the pair matching it. It is pointed at both gateways and talks to whichever one currently holds configuration access, so after a failover it follows the promoted gateway without being told.
Open the workload in the Barbara Panel and edit App Config. The mechanism itself is the same one every Barbara workload uses, described in Application configuration types.
Example
{
"ignition": {
"system": {
"debugLevel": "info",
"applyMode": "enforce",
"reconcileIntervalSeconds": 300
},
"gateways": ["http://gateway:8088", "http://gatewayr:8088"],
"opcuaDevices": [
{
"name": "PLC_Line1",
"driver": "ModbusTcp",
"hostname": "10.20.0.5",
"port": 502
}
],
"opcConnections": [
{
"name": "central-opcua",
"endpointUrl": "opc.tcp://10.20.0.50:62541",
"securityPolicy": "Basic256Sha256",
"securityMode": "SignAndEncrypt",
"authenticationType": "USERNAME",
"username": "barbara",
"password": "secret.centralOpcuaPassword"
}
],
"gatewayNetworkOutgoing": [
{
"name": "central-hub",
"host": "hub.plant.example.com",
"port": 8060,
"useSSL": true
}
]
}
}
gatewayslists both halves of the pair. Keep both entries: that is what lets the configuration follow a failover.opcuaDevicesare the field devices the gateways poll.driveris the Ignition driver name, for exampleModbusTcp.opcConnectionsare OPC UA servers the gateways read from. Apasswordhere is never a literal: it names a secret, as described below.gatewayNetworkOutgoingdeclares a link to another Ignition gateway. WithuseSSLset totrue, Barbara establishes the certificate trust at both ends while the application starts, so you do not approve anything by hand.
What you can declare
| Section | What it configures |
|---|---|
system | How Barbara applies the rest: log level, apply mode, and how often it checks. |
gateways | The gateways this configuration applies to. |
opcuaDevices | Device connections, one per field device. |
opcConnections | Connections to other OPC UA servers. |
opcuaServer | The gateway's own OPC UA server. |
tagProviders | Tag providers, including one that mirrors the tags of another gateway. |
gatewayNetwork | The Gateway Network security settings of the pair. |
gatewayNetworkOutgoing | Links this pair opens to other gateways. |
gatewayNetworkIncoming | Connections from other gateways that the pair accepts. |
databaseConnections | Database connections. |
historianProviders | Where history is written. |
storeAndForwardEngines | The store-and-forward buffer of each database connection. |
Every section, every field and its default value is documented in the application's technical notes on the Barbara Marketplace.
Secrets
A password never goes into Application Config as a literal. Instead you give the passwords once, at install time, in the IGNITION_SECRETS setting of the bbrinit tab, as the base64 of a JSON object of name to value:
{"centralOpcuaPassword": "the-real-password"}
Then refer to each one from Application Config as secret.<name>, exactly as central-opcua does in the example above. The value is stored in the gateway's own secret store, and the gateway configuration only ever holds the reference.
How your changes are applied
Barbara checks the pair against your document every few minutes, and again as soon as you edit it, so a change typically lands within a minute or two.
With applyMode set to enforce, Barbara corrects anything that drifts away from what you declared, including changes made by hand in the Designer. Set it to plan to see what Barbara would change, written to the application log, without touching the gateways.
Barbara only manages what you declare. Resources you created yourself and never named in the document are left alone.
MQTT Transmission is not available in the cluster versions of this application: the module is not installed on the gateways, so declaring MQTT servers or transmitters has no effect. Use the single-node Deploy Ignition application when a site has to publish over MQTT.
Summary
You have deployed a highly available Ignition cluster on Barbara. Traefik provides a single entry point to the active gateway, while Ignition continuously synchronizes the Master and Backup gateways.
If the active gateway becomes unavailable, Traefik redirects traffic in approximately 5–8 seconds, and Ignition completes the promotion in about 30 seconds, allowing the system to continue operating with minimal disruption.
You can monitor the redundancy state at any time from Platform → Redundancy in the Ignition Gateway UI.
From here, declare your device connections, OPC UA connections and Gateway Network links under Application Config and Barbara keeps both gateways matching them, through failovers and rebuilds alike. The same document deploys to the next cluster, which is what makes a site repeatable rather than hand-built.