Red Hat

WildFly 20.0.1 is released!

WildFly 20.0.1.Final is now available for download.

It’s been about a month since the WildFly 20 release, so it’s time for a small bug fix update, WildFly 20.0.1.

The full list of issues resolved in WildFly 20.0.1 is available here. Issues resolved in the WildFly Core 12.0.2 and 12.0.3 releases included with WildFly 20.0.1 are available here and here.

Onward to WildFly 21!


WildFly and Jakarta EE 9

Congratulations to the Jakarta EE community for the recent great progress on Jakarta EE 9!

The Jakarta EE community has been making great strides in its work on Jakarta EE 9, and given today’s Jakarta EE 9 milestone release I wanted to give the WildFly community an update on what’s been going on regarding EE 9 in WildFly and a heads up on what I expect will be happening over the summer and the rest of this year.

As discussed in the Jakarta EE 9 Release Plan, EE 9 is primarily about implementing the necessary change in the Jakarta EE APIs from the javax.* package namespace to the jakarta.* namespace. It isn’t about bringing new functionality to end users; the focus is on providing a platform that all of us in the EE ecosystem can use to adapt to the namespace change, ensuring we’re all in a solid position to take advantage of new features and approaches to doing things that we’d like to see in EE 10.

The WildFly project is an important part of the EE ecosystem, so of course we’re going to participate in this. Besides work from WildFly community members on the Jakarta platform (big shout out to Scott Marlow for his TCK work) and the different specs, there’s been background prototyping work going on exploring how WildFly can provide an EE 9 compatible distribution. That work is now far enough along that it’s time to make it a part of the main WildFly development work.

The javax.* to jakarta.* transition is a big task and it’s going to take a while to percolate through our ecosystem. I don’t think it’s good for WildFly to stop providing new features and fixes to our community while we take this on, so I’d like WildFly’s primary distribution to continue to be based on the EE 8 APIs. I think this should continue to be the case until we begin work toward EE 10.

But we also need to provide an EE 9 server so our community can see what EE 9 will mean to them and so they can use us in their own EE 9 work. So I’d like us to begin producing a tech preview/beta EE 9 variant of WildFly. Ideally there would be at least one very early alpha type milestone over the summer but I don’t expect the first version to appear on the page until some time after the WildFly 21 release, perhaps late September or October. Then another version shortly after the WildFly 22 release, probably in December or early January. Eventually I’d like these to start coming out at the same time as the main WildFly releases.

The main goal of these is to allow people to adapt to the jakarta.* namespace change. However, I would also like them to serve as a bit of a preview for how we see WildFly evolving in the future. For example WildFly 21 will still have the legacy Picketbox-based security as the default security layer, but I’d prefer not to have that layer even be present in the EE 9 variant.

Although I’d like this EE 9 variant to be an evolution from what we have now, and a good way to adapt to the namespace change, it’s important to point out that any EE 10 variant of WildFly may evolve quite significantly from what we’ll be doing with EE 9. There is some uncertainty around how EE 10 will evolve and an expectation that EE 10 and Eclipse MicroProfile alignment will be a key focus, so what we’re doing with EE 9 is likely not going to align fully with our efforts in the future. We are working on getting this notion better codified.

WildFly is a huge codebase, so maintaining two completely distinct flavors of it is not feasible. Furthermore, for a long time at least some of the binaries we ship will have been compiled against EE 8 APIs, with no native EE 9 variant available. To make this work, the EE 9 server would be based on a separate Galleon feature pack from what we use for the main distribution. The large majority of the software artifacts that feature pack references will be the same as what’s in the EE 8 distribution. However, as part of provisioning, any EE 8 content in the server will be transformed (primarily bytecode transformation) to use the EE 9 APIs. Scott Marlow, Richard Opalka and Jean-Francois Denise, with much appreciated assistance from B.J. Hargrave and others on the Eclipse Transformer project, have been making good progress on the needed transformation technology, and Jean-Francois has done well with the needed Galleon tooling. Jean-Francois’s latest POC is able to provision a server that can pass a significant chunk of the WildFly testsuite. That’s a good sign that it’s time for this work to start surfacing in the main WildFly and WildFly Core repos.

Expect to hear more discussion, JIRAs, PRs, etc about this in the coming few weeks as we begin implementing changes in the main code base to make the EE 9 variant more maintainable and as development branches get underway. I’d love to hear your voices!

To be honest, when the need for the javax.* to jakarta.* transition came up last year I was dreading dealing with it, but now I think it will be a lot of fun. Part of the overall goal with what we’ve been doing with Galleon has been to make it easier for users to have the WildFly they want. That rightfully should include truly distinct flavors, not just different subsets of a single flavor. This EE 9 work is going to be a great opportunity for us to make progress on that goal.

Best regards,


Introducing the WildFly MicroProfile Reactive Specifications Feature Pack

I am pleased to announce the 1.0.0.Beta1 release of the MicroProfile Reactive specifications feature pack for WildFly. It offers experimental support for the following MicroProfile specifications, which all focus on the reactive area:

  • MicroProfile Reactive Messaging 1.0 - this is a framework for building event-driven, data streaming and event sourcing applications using CDI. The streams, or channels, can be backed by a variety of messaging technologies. We currently ship connectors for: Apache Kafka, AMQP and MQTT.

  • MicroProfile Reactive Streams Operators 1.0 - Reactive Messaging is build on Reactive Streams. RSO gives you a way to manipulate and handle those streams.

  • MicroProfile Context Propagation 1.0 - The traditional way of propagating state using ThreadLocals does not work well in the reactive world. Async/reactive code often creates a 'pipeline' of code blocks that get executed 'later' - in practice after the method defining them has returned. MicroProfile Context Propagation is there to help you deal with this, so that your deferred code can still for example latch onto the transaction initiated by the calling method.

We are using the SmallRye implementations of each of these specifications.

The source code for the feature pack can be found on GitHub. The README contains links to the specifications, as well as the SmallRye implementations of these and documentation.

Installing the feature pack

We decided to see what the interest is in using these MicroProfile Reactive specifications in WildFly before integrating them into the WildFly code itself, which is why we have shipped this as a Galleon feature pack. This is something that we plan on doing a lot more of in the future for experimental features. Galleon is a tool we have been using internally to compose the server the past several major releases of WildFly.

To install the feature pack, download the latest version of Galleon. At the time of writing this is 4.2.5. Unzip it somewhere, and add its bin/ folder to your path.

Next, save a copy of provision.xml somewhere, and go to that folder in a terminal window. Then run:

$ provision ./provision.xml --dir=my-wildfly

This will take some time the first time you do it since it will download a lot of dependencies from Maven. Once that is done, subsequent attempts will be fast.

What this command does is:

  • Provision a slimmed version (compared to the full download) of WildFly containing the relevant parts for a server running in the cloud. The main README of the project repository contains more information about this part. You can adjust this file to choose other parts of the server you may be interested in.

  • Next it provisions the full contents of the feature pack into our new server instance.

  • The provisioned server will be output in the my-wildfly subdirectory, and can be started via the usual my-wildfly/bin/ command.


A short example of what these specs can do follows. The code snippets are inspired by the Quickstarts, so be sure to try those out!

First we have a method which generates a new price every five seconds:

    private Random random = new Random();

    public Flowable<Integer> generate() {
        return Flowable.interval(5, TimeUnit.SECONDS)
                .map(tick -> random.nextInt(100));

The @Outgoing annotation comes from Reactive Messaging, and specifies that the stream of generated prices will be sent to a channel called 'generated-price'. Channels may be either in-memory, or they may be backed by a messaging provider.

In this case, we have another method (it can be in another class) annotated with @Incoming, using the same 'generated-price' name:

    public double process(int priceInUsd) {
        return priceInUsd * CONVERSION_RATE;

The @Incoming annotation tells it to listen for messages on the generated-price channel. There is a match with the name of the @Outgoing annotation in the previous example so this method will receive all the prices generated by the generate() method. As the name is the same in the two annotations, this becomes an in-memory stream.

The method is also annotated with an @Outgoing annotation so once its conversion has been done, the result is sent to the 'to-kafka' channel.

To map this channel to a Kafka stream, we need some configuration, using MicroProfile Config in a that is part of the deployment:

# Selects the Kafka connector for the 'to-kafka' outgoing stream
# Maps the outgoing stream to the 'prices' Kafka topic
# Adds a serializer to convert the data

Next we create a Publisher that reads this data from Kafka.

    @Channel("from-kafka") Publisher<Double> prices;

This @Channel annotation on a Publisher is conceptually the same as if we had annotated a method with @Incoming("from-kafka") but allows us to do some cool tricks which we will see soon. This is not part of the current Reactive Messaging 1.0 specifaction, but will be part of 1.1. For now it is a SmallRye extension to the specification.

In our that is part of the deployment we configure this channel mapping to the same Kafka stream:

# Selects the Kafka connector for the 'from-kafka' incoming stream
# Maps the incoming stream to the 'prices' Kafka topic
# Adds a deserializer to convert the data

To summarise where we are at so far all the messages which got generated in our generate() methods got sent, via an in memory channel, to our process() method. The process() method did some conversion before sending it to a Kafka topic called 'prices'. Then we listen to that Kafka topic, and are able to publish them from our prices Publisher instance.

Now that we have the converted stream in a Publisher instance we can access it from the non-reactive world, e.g. in a REST endpoint:

    @Produces(MediaType.SERVER_SENT_EVENTS) // denotes that server side events (SSE) will be produced
    @SseElementType(MediaType.TEXT_PLAIN) // denotes that the contained data, within this SSE, is just regular text/plain data
    public Publisher<Double> readThreePrices() {
        // get the next three prices from the price stream
        return ReactiveStreams.fromPublisher(prices)

To keep things simple, we will consider the above simple version of this method first. As we got the stream into a Publisher by using the @Channel annotation, we have a bridge into the 'user world' from the 'reactive world'. Otherwise we would just have a chain of @Outgoing and @Incoming annotated methods (which of course may be also useful in some cases!).

First, we use the MicroProfile Reactive Streams Operators method ReactiveStreams.fromPublisher() to wrap the publisher. We then specify limit(3) - this has the effect that once someone calls this method the stream will terminate after receiving three prices. We call buildRs() to return a new Publisher for those three items. As the messages are every five seconds the readPrices() method will return while our reactive stream is still receiving and re-emitting the three messages.

Next, let’s see how MicroProfile Context Propagation is useful. We will modify the above method, so that each of the three prices get stored to a database

    @PersistenceContext(unitName = "quickstart")
    EntityManager em;

    @Transactional // This method is transactional
    @Produces(MediaType.SERVER_SENT_EVENTS) // denotes that server side events (SSE) will be produced
    @SseElementType(MediaType.TEXT_PLAIN) // denotes that the contained data, within this SSE, is just regular text/plain data
    public Publisher<Double> readThreePrices() {
        // get the next three prices from the price stream
        return ReactiveStreams.fromPublisher(prices)
                .map(price -> {
                    // Context propagation makes this block inherit the transaction of the caller
                    System.out.println("Storing price: " + price);
                    // store each price before we send them
                    Price priceEntity = new Price();
                    // here we are all in the same transaction
                    // thanks to context propagation

                    return price;

First of all we have made the method transactional, so a transaction will be started when entering the method. We then read three prices exactly the same as before, but this time we have an extra call to map(). Inside the map() block, we save each price to a database. Thanks to Context Propagation (which is integrated with Reactive Streams Operators) this happens within the transaction of the readThreePrices() method, although that method will have completed by the time the prices come through.


We’re keen to hear your feedback! Please raise any issues found at

back to top