Skip to content

A OPC UA gateway which gives you access to your OPC UA values via MQTT or GraphQL (HTTP). If you have an OPC UA server in your PLC, or a SCADA system with an OPC UA server, you can query data from there via MQTT and GraphQL (HTTP). In addition, the gateway can also log value changes from OPC UA nodes in an InfluxDB, IoTDB, Kafka, and others.

License

Notifications You must be signed in to change notification settings

chenliuyang1989/automation-gateway

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Frankenstein Automation Gateway

Connect one or more OPC UA servers to the gateway and access the data from the OPC UA servers with a GraphQL or a MQTT client.

Docker images can be found on Docker Hub.

News and Blog posts can be found here.

You can sponsor this project here.

Gateway

Version History

1.20 Cleanup and GraalVM Native Build
1.19 Neo4j Logger to write field values to the graph database
1.18.3 Added MQTT Websocket Option and simple Authentication
1.18.2 Raw value to engineering value conversion for PLC4X driver
1.18.1 Features and bug fixes in PLC4X driver
1.18 Removed Apache Ignite
1.17 Added CrateDB as supported JDBC database for logging
1.16 JDBC Logger to write field values to relational databases
1.15.1 Added BrowsePath to GraphQL OPC UA node (with Full option)
1.15 Nats Logger to write field values to a Nats server
1.14 Fixes and optimizations
1.13 MQTT Logger to write field values to a MQTT Broker
1.12 MQTT Driver with Groovy script transformer (subscribe only)
1.11 Apache Kafka Database Logger
1.10 Apache IoTDB Database Logger
1.9 Apache Ignite as Cluster option and Ignite as Memory-Store
1.8 Upgrade to VertX 4.0.3
1.7 DDS Driver (subscribe and publish)
1.6 Added GraphiQL (http://localhost:4000/graphiql/)
1.5 OPC UA Schemas to GraphQL Schema Importer
1.4 Build GraphQL Schema from OPC UA Schemas
1.3 OPC UA Browsing and fixes
1.2 Simple Polling for PLC4X Driver
1.1 PLC4X Driver
1.0 Initial Version

Build and Run

It needs Java 11 or higher. There is an issue with Java 8 update 292 and bouncycastl encryption. here you can find more information about this issue..

You can open the project in IntelliJ IDEA IDE and build it there or use grade to build it from command line. You can download gradle and build it from command line. Use version 7 or higher of gradle! Check the version of gradle if you install gradle on Linux with your package management system. On some distributions gradle is too old and will throw errors during build phase.

> cd source/app
> gradle build

> export GATEWAY_CONFIG=config.yaml
> gradle run

App is a single program with GraphQL, MQTT and the OPC UA connections in one single program. There is also an option to let it run in clustered mode. Then the modules (app-gateway, app-influxdb, app-plc4x, ...) run in separate processes and form a cluster with a distributed communication bus based on Vert.x. With that option the modules can run on different nodes or they can be deployed in a Kubernetes Cluster. Also some options are only available in clustered mode (like plc4x), but if you want a single app with for example plc4x, then you can also include the plc4x parts in the single app, but you have to code it by your own.

Configuration

See config.yaml in the app directory for an example how to configure the Gateway. You can pass a configuration file name to the program as the first argument or by setting a environment variable GATEWAY_CONFIG. If no argument is given then config.yaml will be used. If there are any questions about the configuration, please get in touch with me.

If you enable GraphiQL (a graphical ui to build and execute GraphQL queries) you can access GraphiQL with http://localhost:4000/graphiql/ ! trailing slash is important !

GraphQLServer:
  Listeners:
    - Port: 4000
      Enabled: true
      LogLevel: INFO
      GraphiQL: true

Example Topics

Note: remove the blanks between the slashes! (just here for better readabilty)

Using the NodeId

opc / test / node / ns=2;s=ExampleDP_Float.ExampleDP_Arg1
opc / test / node / 2 / ExampleDP_Float.ExampleDP_Arg1

Value as JSON with timestamp, quality, ...

opc / test / node:json / ns=2;s=ExampleDP_Float.ExampleDP_Arg1
opc / test / node:json / 2 / ExampleDP_Float.ExampleDP_Arg1

Using the browse path instead of the NodeId

opc / test / path / root-node-id / browse-name / browse-name /...

Wildcard "+" can also be used as a browsename

opc / ua / path:json / ns=1;s=16|Tags / +

"Objects" can be used as root node and will be replace with "i=85"

opc / test / path / Objects / Test / Test00003 / float
opc / test / path / Objects / Test / Test00003 / +

Be careful when using wildcards when there are a lot of nodes, it can lead to a lot of browsing round trips

opc / test / path / Objects / Test / + / float

Enable OPC UA Schema in GraphQL

The GraphQL server can read the OPC UA object schema and convert it to a GraphQL schema. The starting NodeIds can be set to reduce the amount of browsed items. Browsing can take some while if the OPC UA server holds a huge structure of tags!

GraphQLServer:
  Listeners:
    - Port: 4000
      Enabled: true
      LogLevel: INFO                    # ALL | INFO
      GraphiQL: true
      WriteSchemaToFile: false
      Schemas:                          # This systems will be browsed and converted to GraphQL
        - System: "unified"             # Id of OPC UA system, must correlate with the drivers id
          FieldName: BrowseName         # Use "BrowseName" or "DisplayName" as item name in GraphQL
          RootNodes: 
            - ns=2;s=Simulation         # Node will be browsed and added to GraphQL schema 
            - ns=2;s=SimulationMass     # Node will be browsed and added to GraphQL schema          
OpcUaClient:
  - Id: "unified" 
    Enabled: true
    LogLevel: INFO
    EndpointUrl: "opc.tcp://scada-server:4890"
    UpdateEndpointUrl: scada-server
    SecurityPolicyUri: http://opcfoundation.org/UA/SecurityPolicy#None  

Example GraphQL Query with two OPC UA systems:

{
  Systems {
    unified {
      HmiRuntime {
        HMI_RT_5 {
          Structure_instances{
            A1 {
              Velocity { ...Value }
              RefPoint { ...Value }
            }
          }
        }
      }
    }    
    ignition {
      Tag_Providers {
        default {
          Pump_1 {
            flow { ...Value }
            speed { ...Value }
          }
        }
      }
    }
  }
}

fragment Value on Node {
  Value {
    Value
    SourceTime
  }
}

Build Docker Image

You have to build the program before with gradle. Then you can use the shell script docker/build.sh to build a docker image.
docker run --rm --name gateway -p 4000:4000 -p 1883:1883 -v $PWD/config.yaml:/app/config.yaml gateway

C:\Workspace\automation-gateway\source> gradle build
C:\Workspace\automation-gateway\docker> build.bat
C:\Workspace\automation-gateway\docker\examples\hazelcast> docker compose up -d

Using PLC4X

You have to start the app-gateway plus app-plc4x. See config.yaml in each folder of the app. Because some PLC4X drivers / plc's do not support subscriptions we have added a simple polling options. (currently only one polling time per connection).

> cd app-gateway
> gradle run  

> cd app-plc4x
cat config.yaml
Plc4x:
  Drivers:
    - Id: "machine1"
      Enabled: true
      Url: "modbus://127.0.0.1:502"
      Polling:
        Time: 1000 # ms
        Timeout: 900 # ms
        OldNew: true
      WriteTimeout: 100 # ms
      ReadTimeout: 100 # ms
      LogLevel: ALL

> gradle run

Example GraphQL Query:

{
  a: NodeValue(
    Type: Plc,
    System: "mod"
    NodeId: "coil:1"
  ) {
    Value
  }
  b: NodeValue(
    Type: Plc
    System: "mod"
    NodeId: "holding-register:1:INT"
  ) {
    Value
    SourceTime
  }  
}

Example MQTT Topic:

plc/mod/node:json/holding-register:1:INT
plc/mod/node/holding-register:2:INT
plc/mod/node:json/coil:1
plc/mod/node/coil:1

Version History

1.20 Cleanup and GraalVM Native Build

GraalVM Native Image build is now possible, see directory "native". Removed various unused features and upgraded libraries to the latest versions. For the native build it was needed to replace SLF4J logging with the standard Java logging.

Removed Features:

  • Clustering
  • NATS
  • DDS
  • IoTDB: it leads to debug(!) log messages of other components, did not invest time to find out how to get rid of this behaviour.
  • MQTTDriver: it uses Groovy and I could get it running natively compiled, so I had to remove it to get rid of Groovy

Fixed bug: Topic data class now contains two separate fields "path" and "node". Before there was only one field "address". This also fixed a Bug when connecting/disconnecting of topics with wildcards.

1.19 Neo4j Logger to write field values to the graph database

Added Neo4j as an option to log values from OPC UA to the graph database. Additionally the OPC UA node structure can also be replicated to the graph database. This will be done only once at the startup of the Automation Gateway.

Database:
  Logger:
    - Id: neo4j
      Enabled: true
      Type: Neo4j
      Url: bolt://nuc1.rocworks.local:7687
      Username: "neo4j"
      Password: "manager"
      Schemas:
        - System: opc1  # Replicate node structure to the graph database
          RootNodes:
            - "ns=2;s=Demo"  # This node and everything below this node
        - System: winccoa1  # Replicate the nodes starting from "i=85" (Objects) node
      WriteParameters:
        BlockSize: 1000
      Logging:
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Float/+
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Double/+
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Int16/+
        - Topic: opc/winccoa1/path/Objects/PUMP1/#
        - Topic: opc/winccoa1/path/Objects/ExampleDP_Int/#


1.18.3 Added MQTT Websocket Option and simple Authentication

Added the option to enable a Websocket listener for the MQTT server. The Websocket listener is listening on the endpoint "/mqtt". Example MQTT Client Url: "ws://your-host-or-ip/mqtt"

MqttServer:
  Listeners:
    - Id: Mqtt  # Tcp listener without authentication
      Port: 1883

    - Id: MqttWs  # Websocket listener with authentication
      Port: 1884
      Websocket: true  
      Username: system  # If empty, then any username can be used
      Password: manager  # If empty, then any or no password can be used

1.18.2 Raw value to engineering value conversion for PLC4X driver

It is now possible to define Groovy functions to convert raw values to engineering values. Every incoming or outgoing value will be passed through the defined functions. The functions can currently only be defined at the connection level, so every value coming from this PLC connection goes through the conversion functions.

Plc4x:
  Drivers:
    - Id: "niryo"
      Enabled: true
      Url: "modbus://192.168.1.9:5020"
      Polling:
        Time: 100
        Timeout: 90
        OldNew: true
      WriteTimeout: 100
      ReadTimeout: 100
      LogLevel: ALL
      Value:
        Reader: > # Here is the function for incoming values
          def x = value as int;
          return x > 32767 ? 32767 - x : x
        Writer: > # Here is the function for outgoing values
          def x = value as int;
          return x < 0 ? 32767 - x : x

If you need different functions for different addresses then you have to implement a switch/case statement. The functions have two arguments: "address" and "value".

Here is an example where a value conversion is done only for some addresses:

      Value:
        Reader: >
          def x = value as int;
          def xs = ["input-register:1:UINT", "input-register:2:UINT", "input-register:2:UINT"];
          if (xs.contains(address)) {
            println "convert!"
            return x > 32767 ? 32767 - x : x;
          } else {
            println("default")
            return x;
          } 

1.18.1 Features and fixes in PLC4X driver

Auto reconnect to the PLC if the connection is lost.
Read/Write/Poll only if the connection to the PLC is up.
Fixed issue when writing a value fails (stopped the writing thread).
Fixed issue in unsubscribe with polling option (polling item was not removed).
Fixed issue when a client disconnects (unsubscribe with a list of topics failed).
Added new configuration settings for the driver: timeout in ms for write, read, polling.

Plc4x:
  Drivers:
    - Id: "machine1"
      Enabled: true
      Url: "modbus://127.0.0.1:502"
      Polling:
        Time: 1000
        Timeout: 900 # ms
        OldNew: true
      WriteTimeout: 100 # ms
      ReadTimeout: 100 # ms
      LogLevel: INFO

1.18 Removed Apache Ignite

Apache Ignite was removed due to its size and maintenance requirements. We don't know of anyone using the Ignite option, so we decided to remove it.

1.17 Added CrateDB as supported JDBC database for logging

CrateDB is now also supported as JDBC database for logging. If the table (default name "events") does not exists, it will be created partitioned by the month of the source time with four shareds. But you can create the table manually in advance with the settings of your needs.

CREATE TABLE IF NOT EXISTS $sqlTableName (
  "sys" TEXT, 
  "nodeid" TEXT,          
  "sourcetime" TIMESTAMP WITH TIME ZONE,
  "servertime" TIMESTAMP WITH TIME ZONE,
  "sourcetime_month" TIMESTAMP WITH TIME ZONE GENERATED ALWAYS AS date_trunc('month', "sourcetime"),
  "numericvalue" DOUBLE,
  "stringvalue" TEXT,
  "status" TEXT,
  PRIMARY KEY (sourcetime_month, sourcetime, sys, nodeid)
) CLUSTERED INTO 4 SHARDS PARTITIONED BY ("sourcetime_month");   

1.16 JDBC Logger to write field values to relational databases

Added the option to log values to a JDBC compliant relational database. You have to add the JDBC driver to your classpath and set the appropriate JDBC URL path in the configuration file. PostgreSQL, MySQL and Microsoft SQL Server JDBC drivers are already included in the build.gradle file (see lib-jdbc/build.gradle) and also appropriate SQL statements are implemented for those relational databases. If you use other JDBC drivers you can add the driver to the lib-jdbc/build.gradle file as runtime only dependency and you may specify SQL statements for insert and select in the configuration file.

You can specify the table name in the config file with the option "SqlTableName", if you do not specify the table name then "events" will be used as default name.

Specify JDBC drivers in the lib-jdbc/build.gradle file:

    runtimeOnly group: 'org.postgresql', name: 'postgresql', version: '42.2.20'
    runtimeOnly group: 'mysql', name: 'mysql-connector-java', version: '8.0.25'
    runtimeOnly group: 'com.microsoft.sqlserver', name: 'mssql-jdbc', version: '9.2.1.jre11'

Create a table with this structure. For PostgreSQL, MySQL and Microsoft SQL Server the table will be created on startup automatically.

  CREATE TABLE IF NOT EXISTS public.events
  (
      sys character varying(30) NOT NULL,
      nodeid character varying(30) NOT NULL,
      sourcetime timestamp without time zone NOT NULL,
      servertime timestamp without time zone NOT NULL,
      numericvalue numeric,
      stringvalue text,
      status character varying(30) ,
      CONSTRAINT pk_events PRIMARY KEY (system, nodeid, sourcetime)
  )
  TABLESPACE ts_scada;

Configuration of JDBC database logger:

Database:
  Logger:
    - Id: postgres
      Type: Jdbc
      Enabled: true
      Url: jdbc:postgresql://nuc1:5432/scada
      Username: system
      Password: manager
      SqlTableName: events     
      WriteParameters:
        QueueSize: 20000
        BlockSize: 10000
      Logging:
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_SByte/+
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Byte/+

Because the SQL dialect can be slightly different with other databases, you can specify the insert and select SQL statement in the config file:

Database:
  Logger:
    - Id: other
      Type: Jdbc
      Enabled: true
      Url: jdbc:other://nuc1:1111/scada
      Username: system
      Password: manager
      SqlTableName: events     
      SqlInsertStatement: > 
        INSERT INTO events (sys, nodeid, sourcetime, servertime, numericvalue, stringvalue, status)
         VALUES (?, ?, ?, ?, ?, ?, ?)
         ON CONFLICT ON CONSTRAINT PK_EVENTS DO NOTHING  
      SqlQueryStatement: >
        SELECT sourcetime, servertime, numericvalue, stringvalue, status
         FROM events
         WHERE sys = ? AND nodeid = ? AND sourcetime >= ? AND sourcetime <= ? 

1.15 Nats Logger to write field values to a Nats server

Added a Nats Logger to write field values to a Nats server. It is like a database logger, but it writes the values to a configurable Nats server. Any values which get into Frankenstein (OPC UA, PLC4X, DDS, MQTT) by a Driver can be logged to a Nats server. The values are stored in JSON format.

Database:
  Logger:
    - Id: nats1
      Type: Nats
      Enabled: true
      Url: "nats://nuc1:4222"
      WriteParameters:
        QueueSize: 20000
        BlockSize: 10000
      Logging:
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_SByte/+
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Byte/+

1.14 Fixes and optimizations

  • MQTT Driver: Caching of NodeIds in MQTT Driver (topic to NodeId cache).
  • MQTT Server and Driver: max message size can be set in config file (e.g. for Video streaming).
  • GraphQL Define one or more starting/root NodeIds for GraphQL OPC UA schema import.
  • GraphQL: Renamed "NodeIds" to "RootNodes" in Schema section (for OCP UA )
  • Renamed TopicValueDDS to TopicValueJson.

1.13 MQTT Logger to write field values to a MQTT Broker

Added a MQTT Logger to write field values to a MQTT Broker. It is like a database logger, but it writes the values to a configurable MQTT Broker. Any values which get into Frankenstein (OPC UA, PLC4X, DDS, MQTT) by a Driver can be logged to a MQTT Broker. The values are stored in JSON format.

Database:
  Logger:
    - Id: mqtt1
      Type: Mqtt
      Enabled: true
      Host: 192.168.1.169
      Port: 1883
      Ssl: false
      WriteParameters:
        QueueSize: 20000
        BlockSize: 10000
      Logging:
        - Topic: opc/smarthome/path/Meter_Input/WattAct
        - Topic: opc/smarthome/path/Meter_Output/WattAct
        - Topic: opc/smarthome/path/PV/Spot/+

1.12 MQTT Driver with Groovy script transformer (subscribe only)

Added a inital version of a MQTT Driver to get values from a MQTT Broker into Frankenstein. A Groovy script can be used to transform the values to an OPC UA format, so that Frankenstein can be used to log those value to databases. Functionality is currently very limited, only subscribe is implemented.

In this example we transform values of a MQTT Broker from this format: {"TimeMS":1620327963328,"Value":10.277357833719135} to our internal TopicValueOpc format by using a Groovy script and then log some topic values to an InfluxDB.

MqttClient:
  - Id: "mqtt1"
    Enabled: true
    LogLevel: INFO
    Host: 192.168.1.6
    Port: 1883
    Ssl: false
    Value:
      Format: JSON
      Script: >
        return [ 
          className: "TopicValueOpc",
          sourceTime: Instant.ofEpochMilli(source.TimeMS).toString(),
          serverTime: Instant.now().toString(),
          value: source.Value,
          dataTypeId: 0,
          statusCode: 0 ]

Database:
  Logger:
    - Id: influx1
      Type: InfluxDB
      Enabled: true
      Url: http://192.168.1.13:8086
      Database: test
      Username: ""
      Password: ""
      WriteParameters:
        QueueSize: 20000
        BlockSize: 10000
      Logging:
        - Topic: mqtt/mqtt1/path/Meter_Input/WattAct
        - Topic: mqtt/mqtt1/path/Meter_Output/WattAct
        - Topic: mqtt/mqtt1/path/PV/Spot/+          

1.11 Apache Kafka Database Logger

Added Apache Kafka as tag logger option, all incoming value changes of the configured topics will be published to an Apache Kafka Broker. How to can be found here

Database:
  Logger:
    - Id: kafka1
      Type: Kafka
      Servers: server2:9092
      WriteParameters:
        QueueSize: 20000
        BlockSize: 10000
      Logging:
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_SByte/+
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Byte/+

You can also use KSQL to analyze the tag stream. Connect to a ksql cli:

docker exec -ti ksqldb-cli ksql http://ksqldb-server:8088

Create a stream for your logger. Each logger has its own Kafka-Topic. Topic name is equal to the source id of the tag (opc/opc1/...)

CREATE STREAM opc1(
  browsePath VARCHAR KEY, 
  sourceTime VARCHAR, 
  value DOUBLE, 
  statusCode VARCHAR
) WITH (
  KEY_FORMAT='KAFKA',
  KAFKA_TOPIC='opc1', 
  VALUE_FORMAT='JSON',
  TIMESTAMP='sourceTime',TIMESTAMP_FORMAT='yyyy-MM-dd''T''HH:mm:ss[.n]X'
);

Example of a simple quey:

SELECT node, COUNT(*) 
FROM opc1 
WINDOW SESSION (10 SECONDS) 
WHERE node like '%_00' 
GROUP BY node 
EMIT CHANGES;

1.10 Apache IoTDB Database Logger

Added Apache IoTDB as tag logger option.

Database:
  Logger:
   - Id: iotdb1
      Type: IoTDB
      Host: server2
      Port: 6667
      Database: root.scada1
      Username: "root"
      Password: "root"
      WriteParameters:
        QueueSize: 20000
        BlockSize: 10000
      Logging:
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_SByte/+
        - Topic: opc/opc1/path/Objects/Demo/SimulationMass/SimulationMass_Byte/+

1.9 Apache Ignite as Cluster option and Ignite as Memory-Store

Added Apache Ignite as an option for clustering and also to use the Apache Ignite Distributed In Memory Cache for storing last and history values coming from OPC UA or other sources. With that enabled it is possible to do SQL queries on the process values. The cache node stores historical value changes for a defined timerange. So older values are purged on a regular basis (configurable in the configuration file). It is configurable which topics should be stored in the Apache Ignite Cache.

Cache:
  - Id: "global"
    Enabled: true
    LogLevel: INFO
    SqlIndexMaxInlineSize: 1000
    StoreHistoryValues: true
    Systems:
      - SystemType: Opc
        SystemName: "opcua1"
        PurgeEverySeconds: 10
        KeepLastSeconds: 180
      - SystemType: Opc
        SystemName: "opcua2"
        PurgeEverySeconds: 10        
        KeepLastSeconds: 180                                    
    Logging:
      - Topic: opc/opcua1/path/Objects/Demo/SimulationMass/SimulationMass_Double/+
      - Topic: opc/opcua2/path/Objects/Demo/SimulationMass/SimulationMass_Double/+

Updates on the tables are currently not possible. It is implemented that updates on the updatevalue columns should write the values to the source OPC UA server, but unfortunately it leads to unexpected behaviour of the Ignite Cluster.

There is a command line query tool available in the Apache Ignite Distribution. But you can also download the JDBC driver or ODBC drive from the Apache Ignite Website and use any other JDBC/ODBC client tool.

C:\Tools\apache-ignite-2.9.1-bin\bin\sqlline.bat -u jdbc:ignite:thin://192.168.1.18

Example Queries:

select * from global.opcnode where nodeid like '%Mass%'  

select systemname, count(*) 
from global.opcvalue
 group by systemname;   

select systemname, count(*), min(sourcetime), max(sourcetime) 
from global.opcvaluehistory 
group by systemname;  

There are Docker examples available in the docker/examples directory.

C:\Workspace\automation-gateway\source> gradle build
C:\Workspace\automation-gateway\docker> build.bat
C:\Workspace\automation-gateway\docker\examples\ignite> docker compose up -d

1.8 Upgrade to VertX 4.0.3

Upgraded to VertX 4.0.3 and splitted the value type to a base class with subclasses for Opc, Plc and DDS. The app names have been changed, the clustered apps are now named with "cluster". DDS values can now be logged to InfluxDB.

1.7 DDS Driver (subscribe and publish)

Added a first version of DDS support. Currently only MQTT subscribe and publish to DDS topics are functional. It is the app-dds application, the app-gateway must also be up and running.

You need to install OpenDDS and build it with Java support. And you also have to compile your DDS IDL files with Java support. See the ReadMe.txt in the idl directory of app-dds.

Example MQTT Topic:

dds/system-id/path/topic-type-name/topic-name
dds/demo/path/shape/Circle
dds/demo/path/shape/Square

Configuration

DDS:
  Domains:
    - Id: "demo"
      Enabled: true
      LogLevel: ALL
      DCPSConfigFile: rtps.ini
      Domain: 0
      TopicTypes:
        - Id: "shape"
          TopicTypeName: "org.omg.dds.demo.ShapeType"
        - Id: "message"
          TopicTypeName: "org.omg.dds.demo.Message"

1.6 Added GraphiQL (http://localhost:4000/graphiql/)

Added GraphiQL to the Gateway and optionally write the browsed schemas (OPC UA and generated GraphQL scheam) to files.

GraphQLServer:
  Listeners:
    - Port: 4000
      LogLevel: ALL
      GraphiQL: true  # Enable GraphiQL
      WriteSchemaToFile: false  # Write GraphQL Schema to a file
      Schemas:
        - System: ignition
          FieldName: BrowseName # BrowseName | DisplayName
OpcUaClient:
  - Id: "ignition"
    Enabled: true
    BrowseOnStartup: true           

1.5 OPC UA Schemas to GraphQL Schema Importer

Support multiple OPC UA schemas in GraphQL. Be sure that you have set BrowseOnStartup: true for the OPC UA servers which you want to embed in the GraphQL schema. Additionally it can be defined which OPC UA field should be taken as the GraphQL field name: it can be "BrowseName" or "DisplayName". But be careful, the DisplayName must not be unique below a node, so it can lead to an invalid schema.

GraphQLServer:
  Listeners:
    - Port: 4000
      LogLevel: ALL
      Schemas:
        - System: ignition
          FieldName: BrowseName # BrowseName | DisplayName
OpcUaClient:
  - Id: "ignition"
    Enabled: true
    BrowseOnStartup: false 

About

A OPC UA gateway which gives you access to your OPC UA values via MQTT or GraphQL (HTTP). If you have an OPC UA server in your PLC, or a SCADA system with an OPC UA server, you can query data from there via MQTT and GraphQL (HTTP). In addition, the gateway can also log value changes from OPC UA nodes in an InfluxDB, IoTDB, Kafka, and others.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Kotlin 90.1%
  • Java 5.9%
  • Batchfile 2.5%
  • Shell 1.4%
  • Dockerfile 0.1%