Kafka Health Check readiness is always down until It's consumed the first time

Describe the bug Given an application with the following extensions:

• pom.xml:

         <dependency>
            <groupId>io.quarkus</groupId>
            <artifactId>quarkus-smallrye-reactive-messaging-kafka</artifactId>
        </dependency>

        <dependency>
            <groupId>io.quarkus</groupId>
            <artifactId>quarkus-kafka-streams</artifactId>
        </dependency>

That creates a custom topology:

@Produces
    public Topology buildTopology() {
        StreamsBuilder builder = new StreamsBuilder();

        JsonbSerde<LoginAttempt> loginAttemptSerde = new JsonbSerde<>(LoginAttempt.class);
        JsonbSerde<LoginAggregation> loginAggregationSerde = new JsonbSerde<>(LoginAggregation.class);

        builder.stream("from", Consumed.with(Serdes.String(), loginAttemptSerde))
                .groupByKey()
                .windowedBy(TimeWindows.of(Duration.ofSeconds(windowsLoginSec)))
                .aggregate(LoginAggregation::new,
                        (id, value, aggregation) -> aggregation.updateFrom(value),
                        Materialized.<String, LoginAggregation, WindowStore<Bytes, byte[]>> as(LOGIN_AGGREGATION_STORE)
                                .withKeySerde(Serdes.String())
                                .withValueSerde(loginAggregationSerde))
                .toStream()
                .filter((k, v) -> (v.getCode() == UNAUTHORIZED.getStatusCode() || v.getCode() == FORBIDDEN.getStatusCode()))
                .filter((k,v) -> v.getCount() > threshold)
                .to("target");

        return builder.build();
    }

Spite of the application is correctly working, the health check readiness says that the “target” is down:

{
    "status": "DOWN",
    "checks": [
        {
            "name": "SmallRye Reactive Messaging - readiness check",
            "status": "DOWN",
            "data": {
                "login-http-response-values": "[OK]",
                "login-denied": "[KO]"
            }
        },
        {
            "name": "Kafka Streams topics health check",
            "status": "UP",
            "data": {
                "available_topics": "login-http-response-values"
            }
        }
    ]
}

This was working fine prior to 1.12.0.Final.

Expected behavior Health check should be UP.

Actual behavior Health check is DOWN in 1.12.0.Final, 999-SNAPSHOT.

To Reproduce Steps to reproduce the behavior:

1. git clone https://github.com/Sgitario/quarkus-examples
2. cd reproducers/kafka-streams-reactive-messaging
3. mvn clean install -Dquarkus.version=1.12.0.Final It fails because the test checks whether the health check is UP.

If we build the reproducer with mvn clean install -Dquarkus.version=1.11.5.Final, it works:

{
    "status": "UP",
    "checks": [
        {
            "name": "SmallRye Reactive Messaging - readiness check",
            "status": "UP",
            "data": {
                "login-http-response-values": "[OK]",
                "target": "[OK]"
            }
        },
        {
            "name": "Kafka Streams topics health check",
            "status": "UP",
            "data": {
                "available_topics": "login-http-response-values"
            }
        }
    ]
}

• Quarkus version or git rev: 1.12.0.Final and 999-SNAPSHOT

[cescoffier]

@Sgitario I added a note to the migration guide.

[Sgitario]

Yes, until we find a better way to implement the check.

Thanks for the update on docs. As this is a breaking change for users using Kafka on OpenShift/K8S, how should we state this change in the migration guide?