Chapter 2. Architecture of an authoring environment
In Red Hat Decision Manager, the Business Central component provides a web-based interactive user interface for authoring services. The KIE Server component runs the services.
You can also use Business Central to deploy services onto a KIE Server. You can use several KIE Servers to run different services and control the servers from the same Business Central.
Single authoring environment
In a single authoring environment, only one instance of Business Central is running. Multiple users can access its web interface at the same time, however the performance can be limited and there is no failover capability.
Business Central includes a built-in Maven repository that stores the built versions of the services that you develop (KJAR files/artifacts). You can use your continuous integration and continuous deployment (CICD) tools to retrieve these artifacts from the repository and move them as necessary.
Business Central saves the source code in a built-in Git repository, stored in the
.niogit directory. It uses a built-in indexing mechanism to index the assets in your services.
Business Central uses persistent storage for the Maven repository and for the Git repository.
A single authoring environment, by default, includes one KIE Server.
A single authoring environment, by default, uses the controller strategy. Business Central includes the Controller, a component that can manage KIE Servers. When you configure a KIE Server to connect to Business Central, the KIE Server uses a REST API to connect to the Controller. This connection opens a persistent WebSocket. In an OpenShift deployment that uses the controller strategy, each KIE Server is initially configured to connect to the Business Central Controller.
When you use the Business Central user interface to deploy or manage a service on the KIE Server, the KIE Server receives the request through the Controller connection WebSocket. To deploy a service, the KIE Server requests the necessary artifact from the Maven repository that is a part of Business Central.
Client applications use a REST API to use services that run on the KIE Server.
Figure 2.1. Architecture diagram for a single authoring environment
Clustering KIE Servers and using multiple KIE Servers
You can scale a KIE Server pod to run a clustered KIE Server environment.
In a clustered deployment, several instances of the KIE Server run the same services. These servers can connect to the Business Central Controller using the same server ID, so they can receive the same requests from the controller. Red Hat OpenShift Container Platform provides load-balancing between the servers. The services that run on a clustered KIE Server must be stateless, because requests from the same client might be processed by different instances.
You can also deploy several independent KIE Servers to run different services. In this case, the servers connect to the Business Central Controller with different server ID values. You can use the Business Central UI to deploy services to each of the servers.
The optional Smart Router component provides a layer between client applications and KIE Servers. It can be useful if you are using several independent KIE Servers.
The client application can use services running on different KIE Servers, but always connects to the Smart Router. The Smart Router automatically passes the request to the KIE Servers that runs the required service. The Smart Router also enables management of service versions and provides an additional load-balancing layer.
High-availability authoring environment
In a high-availability (HA) authoring environment, the Business Central pod is scaled, so several instances of Business Central are running. Red Hat OpenShift Container Platform provides load balancing for user requests. This environment provides optimal performance for multiple users and supports failover.
Each instance of Business Central includes the Maven repository for the built artifacts and uses the
.niogit Git repository for source code. The instances use shared persistent storage for the repositories. A persistent volume with
ReadWriteMany access is required for this storage.
An instance of Red Hat DataGrid provides indexing of all projects and assets developed in Business Central.
An instance of Red Hat AMQ propagates Java CDI messages between all instances of Business Central. For example, when a new project is created or when an asset is locked or modified on one of the instances, this information is immediately reflected in all other instances.
The controller strategy is not suitable for clustered deployment. In an OpenShift deployment, a high-availability Business Central must manage KIE Servers using the OpenShift startup strategy.
Each KIE Server deployment (which can be scaled) creates a ConfigMap that reflects its current state. The Business Central discovers all KIE Servers by reading their ConfigMaps.
When the user requests a change in KIE Server configuration (for example, deploys or undeploys a service), Business Central initiates a connection to the KIE Server and sends a REST API request. The KIE Server changes the ConfigMap to reflect the new configuration state and then triggers its own redeployment, so that all instances are redeployed and reflect the new configuration.
You can deploy several independent KIE Servers in your OpenShift environment. Each of the KIE Servers has a separate ConfigMap with the necessary configuration. You can scale each of the KIE Servers separately.
You can include Smart Router in the OpenShift deployment.
Figure 2.2. Architecture diagram for a high-availability authoring environment