Drools + Quarkus + Runtime metrics

Description

This example demonstrates how to enable and consume the runtime metrics monitoring feature in Kogito.

Prerequisites

You will need:

• Java 11+ installed
• Environment variable JAVA_HOME set accordingly
• Maven 3.8.1+ installed
• Docker 19+ (only if you want to run the integration tests and/or you want to use the docker-compose script provided in this example).

How to enable the feature

Import the following dependency in your pom.xml:

<dependency>
 <groupId>org.kie.kogito</groupId>
 <artifactId>kogito-addons-quarkus-monitoring-prometheus</artifactId>
</dependency>

Architecture

Once you compile your Quarkus project, a dashboard for each available endpoint will be generated under the path target/classes/META-INF/resources/monitoring/dashboards/. You can then inject those grafana dashboards during the deployment of the grafana instance.

The use case is summarized in the following schema:

To summarize, the kogito app will expose by default an endpoint /metrics with the prometheus variables, and a prometheus instance will simply fetch the data from there.

Dashboards

Kogito currently exports two types of dashboards depending on the model used on the endpoint:

1. Operational dashboard: this dashboard is generated for DMN and DRL endpoints and it contains
   • Total number of requests on the endpoint.
   • Average per minute of the number of requests on the endpoint.
   • Quantiles on the elapsed time to evaluate the requests.
   • Exception details.
2. Domain specific dashboard: currently this dashboard is exported only for DMN endpoints. In particular, the domain specific dashboard contains a graph for each type of decision in the DMN model. At the moment, only the built-in types number, string and boolean are supported:
   • if the output of the decision is a number, the graph contains the quantiles for that metric (on a sliding window of 3 minutes).
   • If the output is a boolean or a string, the graph contains the number of occurrences for each output (10 minutes average).

You can use these default dashboards, or you can personalize them and use your custom dashboards.

Custom dashboards

User may add custom dashboards defined as json files. Format specification and details are provided in the Official documentation page. There are some conventions to be followed to successfully add such dashboards inside a Kogito project:

1. dashboard files must be stored inside /src/main/resources/META-INF/dashboards directory
2. dashboard file names must start with domain-dashoboard (for domain specific dashboards) or operational-dashboard (for operational ones)
3. dashboard file names must end with .json
4. dashboard file names should not conflict with auto-generated ones
5. the title attribute of custom dashboards should not conflict with auto-generated ones.

Custom dashboards will be available in the grafana panel, together with auto-generated ones.

Compile and Run in Local Dev Mode

It is possible to use docker-compose to demonstrate how to inject the generated dashboards in the volume of the grafana container:

1. Run mvn clean package to build the project and generate dashboards. A docker image tagged org.kie.kogito.examples/dmn-drools-quarkus-metrics-example:1.0 will be built (docker must be installed on your system).
2. Run docker-compose up to start the applications.

The volumes of the grafana container are properly set in the docker-compose.yml file, so that the dashboards are properly loaded at startup.

Example Usage

Once the service is up and running, you can use the following example to interact with the service.

POST /hello

Post “hello”:

curl -H "Content-Type: application/json" -X POST -d '{"strings":["world"]}' http://localhost:8080/hello

the service will return ["hello", "world"]

POST /LoanEligibility

Post:

curl -X POST 'http://localhost:8080/LoanEligibility' -H 'Content-Type: application/json' \
    -d '{
        "Client": {"age": 43,"salary": 1950, "existing payments": 100},
        "Loan": {"duration": 15,"installment": 180}, 
        "SupremeDirector" : "Yes", 
        "Bribe": 1000
    }'

the service will return the decision results.

If you are using the docker-compose script we provided, go to localhost:3000 and have a look at your dashboards.