Tag Archives: microservices

Docker, Kubernetes, and Microservices Replay from Devoxx 2015

Devoxx 2015 BE Arun

Java gives us Write Once Run Anywhere (WORA) because of the common abstraction provided by Java Virtual Machine. The binary byte code produced by Java is understood by the JVM running on multiple operating systems. This allows Java code to run on any operating system. But deploying such applications typically requires to tune the JVM, setup application server, install appropriate database drivers, other similar configuration. Docker nicely complements WORA by defining a way to package Java applications, and include all the configuration in a an easy to describe format. This can be called as Package Once Deploy Anywhere, or PODA.

Docker PODA

Do you want to learn more about how Docker helps with PODA?
How do you create multi-container applications using Docker Compose?
How do you deploy this multi-container application on multiple hosts running in a Docker Swarm cluster?
What new features Docker introduced in 1.9 release – particularly multi-host networking and persistent storage?
How do you deploy Docker containers using Kubernetes?
How does Kubernetes help with scaling applications?
How does Docker Swarm compare with Kubernetes?

I had the privilege of delivering a 3-hour university on Docker and Kubernetes followed by a 3-hour hands-on lab. The replay of the university session is now available:

Slides for the university are available at: github.com/javaee-samples/docker-java/tree/master/slides

The contents from the hands-on lab are available at bit.ly/dockerlab.

In addition, I also gave a 3-hour university on Refactor your Java EE Applications using Microservices and Containers. This session explained:

  • Basic characteristics of Microservices
  • Showed a simple shopping cart monolithic application, and how it was refactored to multiple microservices (github.com/arun-gupta/microservices)
  • Talked about transactions, event sourcing and CQRS
  • How KumuluzEE was used to create standalone JARs
  • Explained how such microservices can be deployed using Docker

Slides from this session are available at: github.com/arun-gupta/microservices/tree/master/slides

Replay is available at:

The response on twitter was quite positive:

And in case you are interested, watch a quick interview with Voxxed team:



Docker and Kubernetes Workshops in Fall 2015

Docker and Kubernetes workshops is going to 4 continents and 9 countries this Fall!

Lets talk about:

  • Get started with Docker and Kubernetes for packaging your applications
  • Microservices using Docker and Kubernetes
  • Clustering architectures
  • Migrating existing applications to Docker and Kubernetes
  • Tooling
  • Debugging tips

I’ll share some of what I know and will learn a lot more from you!

Here is the complete circuit so far:

 Sep 9 -10  javazone-2015
 Sep 15  goto-london-2015
Sep 17 redhat-forum-london-2015
Sep 29 Red Hat Forums, Argentina
 Oct 2  codestars-summit-2015
 Oct 24 – 29  javaone-logo
 Nov 5  Drukwerk (tentative)
 Nov 7  javaday-kiev-2015
 Nov 9 – 13  devoxx-be-2015
 Nov 16 – 18  devoxx-morocco-2015
 Nov 18 – 22  buildstuff-2015

Where will I see you?

Would you like to run with me at any of these events? 5k, 10k, 10mile, half marathon, marathon … you pick the distance and we run together!


Getting Started with ELK Stack on WildFly

Your typical business application would consist of a variety of servers such as WildFly, MySQL, Apache, ActiveMQ, and others. They each have a log format, with minimal to no consistency across them. The log statement typically consist of some sort of timestamp (could be widely varied) and some text information. Logs could be multi-line. If you are running a cluster of servers then these logs are decentralized, in different directories.

How do you aggregate these logs? Provide a consistent visualization over them? Make this data available to business users?

This blog will:

  • Introduce ELK stack
  • Explain how to start it
  • Start a WildFly instance to send log messages to the ELK stack (Logstash)
  • View the messages using ELK stack (Kibana)

What is ELK Stack?

ELK stack provides a powerful platform to index, search and analyze your data. It uses  Logstash for log aggregation, Elasticsearch for searching, and Kibana for visualizing and analyzing data. In short, ELK stack:

  • Collect logs and events data (Logstash)
  • Make it searchable in fast and meaningful ways (Elasticsearch)
  • Use powerful analytics to summarize data across many dimensions (Kibana)


Logstash is a flexible, open source data collection, enrichment, and transportation pipeline.


Elasticsearch is a distributed, open source search and analytics engine, designed for horizontal scalability, reliability, and easy management.


Kibana is an open source data visualization platform that allows you to interact with your data through stunning, powerful graphics.

How does ELK Stack work?

Logstash can collect logs from a variety of sources (using input plugins), process the data into a common format using filters, and stream data to a variety of sources (using output plugins). Multiple filters can be chained to parse the data into a common format. Together, they build a Logstash Processing Pipeline.

Logstash Processing Pipeline

Inputs and outputs support codecs that enable you to encode or decode the data as it enters or exits the pipeline without having to use a separate filter.

Logstash can then store the data in Elasticsearch and Kibana provides a visualization of that data. Here is a sample pipeline that can collect logs from different servers and run it through the ELK stack.

ELK Stack

Start ELK Stack

You can download individual components of ELK stack and start that way. There is plenty of advise on how to configure these components. But I like to start with a KISS, and Docker makes it easy to KISS!

All the source code on this blog is at github.com/arun-gupta/elk.

  1. Clone the repo:
  2. Run the ELK stack:
    This will use the pre-built Elasticsearch, Logstack, and Kibana images. It is built upon the work done in github.com/nathanleclaire/elk.

    docker ps will show the output as:

    It shows all the containers running.

WildFly and ELK

James (@the_jamezp) blogged about Centralized Logging for WildFly with ELK Stack. The blog explains how to configure WildFly to send log messages to Logstash. It uses the highly modular nature of WildFly to install jboss-logmanager-ext library and install it as a module. The configured logmanager includes @timestamp field to the log messages sent to logstash. These log messages are then sent to Elasticsearch.

Instead of following the steps, lets Docker KISS and use a pre-configured image to get you started.

Start the image as:

Make sure to substitute <DOCKER_HOST_IP> with the IP address of the host where your Docker host is running. This can be easily found using docker-machine ip <MACHINE_NAME>.

View Logs using ELK Stack

Kibana runs on an embedded nginx and is configured to run on port 80 in docker-compose.yml. Lets view the logs using that.

  1. Access http://<DOCKER_HOST_IP> in your machine and it should show the default page as:ELK Stack WildFly PatternThe @timestamp field was created by logmanager configured in WildFly.
  2. Click on Create to create an index pattern and select Discover tab to view the logs as:ELK Stack WildFly Output

Try connecting other sources and enjoy the power of distributed consolidated by ELK!

Some more references …

Distributed logging and visualization is a critical component in a microservices world where multiple services would come and go at a given time. A future blog will show how to use ELK stack with a microservices architecture based application.


Microservices and DevOps Journey at Wix

Wix.com started their journey on DevOps and Microservices about two years ago and recently switched from a monolithic application to a microservices-based application. Yes, it took them full two years to complete the transition from monolith to microservices!

I got connected with Aviran Mordo (@aviranm), head of backend engineering at Wix on twitter.

They migrated to microservices because the “system could not scale” and the requirements for functional components were varied. The journey took their WAR-based deployment on Tomcat to fat JAR with embedded Jetty. On a side note, take a look at WildFly Swarm if you are interested in a similar approach for Java EE applications.

Video Interview

I discussed some points with him about this journey and you can watch the same too.

In this discussion, you’ll learn:

  • Why Continuous Delivery and DevOps are important requirements for microservices?
  • How they migrated from a big monolith to smaller monoliths and then a full-blown microservices architecture
  • How database referential integrity constraints were moved from database to application?
  • “micro” in microservices refers to the area of responsibility, nothing to do with LOC
  • Neither REST nor messaging was used for communication between different services. Which protocol was used? JSON-RPC
  • How do services register and discover each other? Is that required during early phases?
  • Why YAGNI and KISS are important?
  • Chef for configuration management and how to make it accessible for massive deployments
  • TeamCity for CI
  • Is 100% automation a requirement? Can 100% automation be achieved? Learn about Petri, Wix’s open source framework for A/B testing
  • Relevance of hybrid cloud (Google, Amazon, Private data center) and redundancy
  • Hardest part of migrating from monolith to microservice
  • How much code was repurposed during refactoring?
  • Where was the most effort spent during the two years of migration?
  • Distributed transactions
  • What was the biggest challenge in DevOps journey? Look out for a nice story towards the end that could be motivating for your team as well 😉

Additional Material

Watch the slides from DevoxxUK:

You can also learn more about their architecture in Scaling Wix to 60m Users.


ZooKeeper for Microservice Registration and Discovery

In a microservice world, multiple services are typically distributed in a PaaS environment. Immutable infrastructure is provided by containers or immutable VM images. Services may scale up and down based upon certain pre-defined metrics. Exact address of the service may not be known until the service is deployed and ready to be used.

This dynamic nature of service endpoint address is handled by service registration and discovery. In this, each service registers with a broker and provide more details about itself, such as the endpoint address. Other consumer services then queries the broker to find out the location of a service and invoke it. There are several ways to register and query services such as ZooKeeper, etcd, consul, Kubernetes, Netflix Eureka and others.

Monolithic to Microservice Refactoring showed how to refactor an existing monolith to a microservice-based application. User, Catalog, and Order service URIs were defined statically. This blog will show how to register and discover microservices using ZooKeeper.

Many thanks to Ioannis Canellos (@iocanel) for all the ZooKeeper hacking!

What is ZooKeeper?

Apache ZooKeeperZooKeeper is an Apache project and provides a distributed, eventually consistent hierarchical configuration store.


ZooKeeper is a centralized service for maintaining configuration information, naming, providing distributed synchronization, and providing group services. All of these kinds of services are used in some form or another by distributed applications.

So a service can register with ZooKeeper using a logical name, and the configuration information can contain the URI endpoint. It can consists of other details as well, such as QoS.

Apache Curator ZooKeeper has a steep learning curve as explained in Apache ZooKeeper Made Simpler with Curator. So, instead of using ZooKeeper directly, this blog will use Apache Curator.

Curator n ˈkyoor͝ˌātər: a keeper or custodian of a museum or other collection – A ZooKeeper Keeper.

Apache Curator has several components, and this blog will use the Framework:

The Curator Framework is a high-level API that greatly simplifies using ZooKeeper. It adds many features that build on ZooKeeper and handles the complexity of managing connections to the ZooKeeper cluster and retrying operations.

ZooKeeper Concepts

ZooKeeper Overview provides a great overview of the main concepts. Here are some of the relevant ones:

  • Znodes: ZooKeeper stores data in a shared hierarchical namespace that is organized like a standard filesystem. The name space consists of data registers – called znodes, in ZooKeeper parlance – and these are similar to files and directories.
  • Node name: Every node in ZooKeeper’s name space is identified by a path. Exact name of a node is a sequence of path elements separated by a slash (/).
  • Client/Server: Clients connect to a single ZooKeeper server. The client maintains a TCP connection through which it sends requests, gets responses, gets watch events, and sends heart beats. If the TCP connection to the server breaks, the client will connect to a different server.
  • Configuration data: Each node in a ZooKeeper namespace can have data associated with it as well as children. ZooKeeper was originally designed to store coordination data, so the data stored at each node is usually small, in less than KB range).
  • Ensemble: ZooKeeper itself is intended to be replicated over a sets of hosts called an ensemble. The servers that make up the ZooKeeper service must all know about each other.
  • Watches: ZooKeeper supports the concept of watches. Clients can set a watch on a znode. A watch will be triggered and removed when the znode changes.

ZooKeeper is a CP system with regards to CAP theorem. This means if there is a partition failure, it will be consistent but not available. This can lead to problems that are explained in Eureka! Why You Shouldn’t Use ZooKeeper for Service Discovery.

Nevertheless, ZooKeeper is one of the most popular service discovery mechanisms used in microservices world.

Lets get started!

Start ZooKeeper

  1. Start a ZooKeeper instance in a Docker container:
  2. Verify ZooKeeper instance by using telnet as:
    Type the command “ruok” to verify that the server is running in a non-error state.The server will respond with “imok” if it is running:
    Otherwise it will not respond at all. ZooKeeper has other similar four-letter commands.

Service Registration and Discovery

Each service, User, Catalog, and Order in our case, has an eagerly initialized bean that registers and unregisters the service as part of lifecycle initialization methods as. Here is the code from CatalogService:

The code is pretty simple, it injects ServiceRegistry class, with @ZooKeeperRegistry qualifier. This is then used to register and unregister the service. Multiple URIs, one each for a stateless service, can be registered under the same logical name.

At this time, the qualifier comes from another maven module. A cleaner Java EE way would be to move the @ZooKeeperRegistry qualifier to a CDI extension (#20). And when this qualifier when specified on any REST endpoint will register the service with ZooKeeper (#22). For now, service endpoint URI is hardcoded as well (#24).

What does ZooKeeper class look like?

  1. ZooKeeper class uses constructor injection and hardcoding IP address and port (#23):
    It does the following tasks:

    1. Loads ZooKeeper’s host/port from a properties file
    2. Initializes Curator framework and starts it
    3. Initializes a hashmap to store the URI name to zNode mapping. This node is deleted later to unregister the service.
  2. Service registration is done using registerService method as:
    Code is pretty straight forward:

    1. Create a parent zNode, if needed
    2. Create an ephemeral and sequential node
    3. Add metadata, including URI, to this node
  3. Service discovery is done using discover method as:
    Again, simple code:

    1. Find all children for the path registered for the service
    2. Get metadata associated with this node, URI in our case, and return.The first such node is returned in this case. Different QoS parameters can be attached to the configuration data. This will allow to return the appropriate service endpoint.

Read ZooKeeper Javadocs for API.

ZooKeeper watches can be setup to inform the client about the lifecycle of the service (#27). ZooKeeper path caches can provide an optimized implementation of the children nodes (#28).

Multiple Service Discovery Implementations

Our shopping cart application has two two service discovery implementationsServiceDisccoveryStatic and ServiceDiscoveryZooKeeper. The first one has all the service URIs defined statically, and the other one retrieves them from ZooKeeper.

Other means to register and discover can be easily added by creating a new package in services module and implementing ServiceRegistry interface. For example, Snoop, etcd, Consul, and Kubernetes. Feel free to send a PR for any of those.

Run Application

  1. Make sure the ZooKeeper image is running as explained earlier.
  2. Download and run WildFly:
  3. Deploy the application:
  4. Access the application at localhost:8080/everest-web/. Learn more about the application and different components in Monolithic to Microservices Refactoring for Java EE Applications blog.


Monolithic to Microservices Refactoring for Java EE Applications

Have you ever wondered what does it take to refactor an existing Java EE monolithic application to a microservices-based one?

This blog explains how a trivial shopping cart example was converted to microservices-based application, and what are some of the concerns around it. The complete code base for monolithic and microservices-based application is at: github.com/arun-gupta/microservices.

Read on for full glory!

Java EE Monolith

A Java EE monolithic application is typically defined as a WAR or an EAR archive. The entire functionality for the application is packaged in a single unit. For example, an online shopping cart may consist of User, Catalog, and Order functionalities. All web pages are in root of the application, all corresponding Java classes are in the WEB-INF/classes directory, resources in WEB-INF/classes/META-INF directory.

Lets assume that your monolith is not designed as a distributed big ball of mud and the application is built following good software architecture. Some of the common rules are:

  • Separation of concerns, possibly using Model-View-Controller
  • High cohesion and low coupling using well-defined APIs
  • Don’t Repeat Yourself (DRY)
  • Interfaces/APIs and implementations are separate, and following Law of Demeter. Classes don’t call other classes directly because they happen to be in the same archive
  • Using Domain Driven Design to keep objects related to a domain/component together
  • YAGNI or You Aren’t Going to Need It: Don’t build something that you don’t need now

Here is how a trivial shopping cart monolithic WAR archive might look like:

Java EE Monolithic

This monolithic application has:

  • Web pages, such as .xhtml files, for User, Catalog, and Order component, packaged in the root of the archive. Any CSS and JavaScript resources that are shared across different webpages are also packaged with these pages.
  • Classes for the three components are in separate packages in WEB-INF/classes directory. Any utility/common classes used by multiple classes are packed here as well.
  • Configuration files for each component are packaged inWEB-INF/classes/META-INF directory. Any config files for the application, such as persistence.xml and load.sql to connect and populate the data store respectively, are also packaged here.

It has the usual advantages of well-known architecture, IDE-friendly, easy sharing, simplified testing, easy deployment, and others. But also comes with disadvantages such as limited agility, obstacle for continuous delivery, “stuck” with a technology stack, growing technical debt, and others.

Even though microservices are all the raze these days, but monoliths are not bad. Even those that are not working for you may not benefit much, or immediately, from moving to microservices. Other approaches, such as just better software engineering and architecture, may help. Microservices is neither a free lunch or a silver bullet and requires significant investment to be successful such as service discovery, service replication, service monitoring, containers, PaaS, resiliency, and a lot more.

don’t even consider microservices unless you have a system that’s too complex to manage as a monolith.

Microservice Architecture for Java EE

Alright, I’ve heard about all of that but would like to see a before/after, i.e. how a monolith code base and how a refactored microservice codebase looks like.

First, lets look at the overall architecture:

Java EE Microservices

The key pieces in this architecture are:

  • Application should be functionally decomposed where User, Order, and Catalog components are packaged as separate WAR files. Each WAR file should have the relevant web pages (#15), classes, and configuration files required for that component.
    • Java EE is used to implement each component but there is no long term commitment to the stack as different components talk to each other using a well-defined API (#14).
    • Different classes in this component belong to the same domain and so the code is easier to write and maintain. The underlying stack can also change, possibly keeping technical debt to a minimum.
  • Each archive has its own database, i.e. no sharing of data stores. This allows each microservice to evolve and choose whatever type of datastore – relational, NoSQL, flat file, in-memory or some thing else – is most appropriate.
  • Each component will register with a Service Registry. This is required because multiple stateless instances of each service might be running at a given time and their exact endpoint location will be known only at the runtime (#17).Netflix Eureka, EtcdZookeeper are some options in this space (more details).
  • If components need to talk to each other, which is quite common, then they would do so using a pre-defined API. REST for synchronous or Pub/Sub for asynchronous communication are the common means to achieve this.In our case, Order component discovers User and Catalog service and talks to them using REST API.
  • Client interaction for the application is defined in another application, Shopping Cart UI in our case. This application mostly discover the services from Service Registry and compose them together. It should mostly be a dumb proxy where the UI pages of different components are invoked to show the interface (#18).A common look-and-feel can be achieved by providing a standard CSS/JavaScript resources.

This application is fairly trivial but at least highlights some basic architectural differences.

Monolith vs Microservice

Some of the statistics for the monolith and microservices-based applications are compared below:

Characteristic Monolith Microservice
Number of archives  1  5

  • Contracts (JAR, ~4 KB)
  • Order (WAR, ~7 KB)
  • User (WAR, ~6 KB)
  • Catalog (WAR, ~8 KB)
  • Web UI (WAR, 27 KB)
Web pages  8  8 (see below)
Configuration Files  4

  • web.xml
  • template.xhtml
  • persistence.xml
  • load.sql
 3 per archive

  • persistence.xml
  • load.sql
  • web.xml
Class files  12 26

  • Service registration for each archive
  • Service discovery classes
  • Application class for each archive
Total archive size  24 KB  ~52 KB (total)

Code base for the monolithic application is at: github.com/arun-gupta/microservices/tree/master/monolith/everest

Code base for the microservices-enabled application is at: github.com/arun-gupta/microservices/tree/master/microservice

Issues and TODOs

Here are the issues encountered, and TODOs, during refactoring of the monolith to a microservices-based application:

  • Java EE already enables functional decomposition of an application using EAR packaging. Each component of an application can be packaged as a WAR file and be bundled within an EAR file. They can even share resources that way. Now that is not a true microservices way, but this could be an interim step to get you started. However, be aware that@FlowScoped bean is not activated in an EAR correctly (WFLY-4565).
  • Extract all template files using JSF Resource Library Templates.
    • All web pages are currently in everest module but they should live in each component instead (#15).
    • Resource Library Template should be deployed at a central location as opposed to packaged with each WAR file (#16).
  • Breakup monolithic database into multiple databases require separate persistence.xml and DDL/DML scripts for each application. Similarly, migration scripts, such as using Flyway, would need to be created accordingly.
  • A REST interface for all components, that need to be accessed by another one, had to be created.
  • UI is still in a single web application. This should instead be included in the decomposed WAR (#15) and then composed again in the dumb proxy. Does that smell like portlets?
  • Deploy the multiple WAR files in a PaaS (#12)
  • Each microservice should be easily deployable in a container (#6)

Here is the complete list of classes for the monolithic application:

Here is the complete list of classes for the microservices-based application:

Once again, the complete code base is at github.com/arun-gupta/microservices.

Future Topics

Some of the future topics in this series would be:

  • Are containers required for microservices?
  • How do I deploy multiple microservices using containers?
  • How can all these services be easily monitored?
  • A/B Testing
  • Continuous Deployment using microservices and containers

What else would you like to see?



Microservices at Red Hat Summit 2015

RedHat Summit Logo

Red Hat Summit is just a few days away. You’ll get the opportunity to learn a plethora of technologies and products in 277 sessions by 402 speakers. Complete agenda shows the list of all sessions, hands-on labs, tracks, and everything else.

When? Jun 21-25, 2015 (DevNation) and Jun 23-26, 2015 (Summit)
Where? Hynes Convention Center, Boston, MA
AgendaSummit and DevNation
Register: Summit and DevNation

Here is the list of 12 sessions that are talking about Microservices:

You’ll learn about:

  • Microservices technical and business aspects
  • Development, Testing, Production of microservices
  • Relationship with PaaS, Containers, and DevOps, Camel, Fabric8

Of course, you can catch any of the speakers in the hallway. You can also stop by at any of the booths in the exhibitor hall, and talk to us about what Red Hat offers for microservices.

I’m giving a few sessions as well, and really looking forward to see you there!


Gilt and Microservices: Why and How

Gilt.com provides instant insider access to top designer labels, at up to 70% off retail. The company went through growing pains when the monolithic application was not able to handle the needs of growing business demand. They looked at microservices for a scalable solution and are one of the live examples of how they are using it effectively to keep up with the velocity and agility of the business.

  • What was the need to refactor monolithic application to microservice?
  • What is the technology stack?
  • Is JVM relevant?
  • Relevance of containers
  • Is DevOps a requirement for microservices?
  • Should the microservices be fully contained or can the data be shared?

Watch the answer to these questions in this interview with Adrian Trenaman (@adrian_trenaman), software engineer at Gilt:

Some other details are at:

Many thanks to Daniel Bryant (@danielbryantuk) for the introduction!

Cloud, Devops, Microservices Track Program Committee at JavaOne 2015


JavaOne 2015 Program Committee wheels are churning to make sure we provide the best content for you this year. There are a total of 9 tracks covering the entire Java landscape and all the track leads and program committee members are reviewing and voting upon the submissions. This is not an easy task especially when there are a lot of submission and the quality of submissions is top notch.

Here is the list of program committee members for Cloud and DevOps track:




 James Turnbull

And I am (@arungupta) leading the track along with @brunoborges!


Complete list of all the program committee members across all the tracks is published here.

Many thanks for all the wonderful submissions, and to all the program committee members for reviewing the proposals!

Just to give you an idea, here is the tag cloud generated from the titles of all submissions for Cloud and DevOps track:


JavaOne 2015 Title Tag Cloud for Cloud/DevOps Track


And a tag cloud from all the abstracts of the same track:

JavaOne 2015 Abstract Tag Cloud for Cloud/DevOps Track


And in case you are wondering, here is the tag cloud of titles from submissions across all the tracks:

JavaOne 2015 Tag Cloud Title


And the same for submissions across all the tracks:

JavaOne 2015 Tag Cloud Abstract


Stay tuned, we are working hard to make sure to provide an excellent content at JavaOne!

Here are a couple of additional links:

And, you can always find all the details at oracle.com/javaone.

WildFly Swarm: Building Microservices with Java EE

Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away

Antoine de Saint-Exupery

This quote by the French writer Antoine de Saint-Exupery was made to substantiate that often less is more. This is true for architect, artist, designer, writer, running, software developer, or in any other profession. Simplicity, minimalism, cutting down the cruft always goes a long way and has several advantages as opposed to something bloated.

What is WildFly Swarm?

WildFly is a light weight, flexible, feature rich, Java EE 7 compliant application server. WildFly 9 even introduced a 27MB Servlet-only distribution. These are very well suited for your enterprise and web applications.

WildFly Swarm takes the notch a bit higher. From the announcement:

WildFly Swarm is a new sidecar project supporting WildFly 9.x to enable deconstructing the WildFly AS and pasting just enough of it back together with your application to create a self-contained executable jar.

The typical application development model for a Java EE application is to create an EAR or WAR archive and deploy it in an application server. All the dependencies, such as Java EE implementations are packaged in the application server and provide the functionality required by the application classes. Multiple archives can be deployed and they all share the same libraries. This is a well understood model and have been used over the past several years.

WildFly Swarm turns the table where it creates a “fat jar” that has all the dependencies packaged in a JAR file. This includes a minimalist version of WildFly, any required dependencies, and of course, the application code itself. The application can simply be run using java -jar.

Each fat jar could possibly be a microservice which can then independently upgrade, replace, or scale. Each fat jar would typically follow single responsibility principle and thus will have only the required dependencies packaged. Each JAR can use polyglot persistence, and use only the persistence mechanism that is required.

Show me the code!

A Java EE application can be packaged as WildFly Swarm fat jar by adding a Maven dependency and a plugin. Complete source code for a simple JAX-RS sample is available at github.com/arun-gupta/wildfly-samples/tree/master/swarm.

WildFly Swarm Maven Dependency

Add the following Maven dependency in pom.xml:

WildFly Swarm Maven Plugin

Add the following Maven plugin in pom.xml:

Create WildFly Swarm Fat Jar

The fat jar can be easily created by invoking the standard Maven target:

This generates a JAR file using the usual Maven conventions, and appends -swarm at the end. The generated WAR file name in our sample is swarm-1.0-SNAPSHOT-swarm.jar.

The generated WAR file is ~30MB, has 134 JARs (all in m2repo directory), and 211 classes. The application code is bundled in app/swarm-1.0-SNAPSHOT.war.

Run WildFly Swarm Fat Jar

This far jar can be run as:

The response can be verified as:

WildFly Swarm Release blog refers to lots of blogs about Servlet, JAX-RS with ShrinkWrap, DataSource via Deployment, Messaging and JAX-RS, and much more.

WildFly Swarm Next Steps

This is only 1.0.0.Alpha1 release so feel free to try out samples and give us feedback by filing an issue.

You have the power of all WildFly subsystems, and can even create embeddable Java EE container as shown in the release blog:

Subsequent blogs will show how a microservice can be easily created using WildFly Swarm.

WildFly Swarm Stay Connected

You can keep up with the project through the WildFly HipChat room, @wildflyswarm on Twitter, or through GitHub Issues.

Microservice Design Patterns

The main characteristics of a microservices-based application are defined in Microservices, Monoliths, and NoOps.  They are functional decomposition or domain-driven design, well-defined interfaces, explicitly published interface, single responsibility principle, and potentially polyglot. Each service is fully autonomous and full-stack. Thus changing a service implementation has no impact to other services as they communicate using well-defined interfaces. There are several advantages of such an application, but its not a free lunch and requires a significant effort in NoOps.

But lets say you understand the required effort, or at least some pieces of it, that is required to build such an application and willing to take a jump. What do you do? What is your approach for architecting such applications? Are there any design patterns on how these microservices work with each other?


Functional decomposition of your application and the team is the key to building a successful microservices architecture. This allows you to achieve loose coupling (REST interfaces) and high cohesion (multiple services can compose with each other to define higher level services or application).

Verb (e.g. Checkout) or Nouns (Product) of your application are one of the effective ways to achieve decomposition of your existing application. For example, product, catalog, and checkout can be three separate microservices and then work with each other to provide a complete shopping cart experience.

Functional decomposition gives the agility, flexibility, scalability, and other *ilities but the business goal is still to create the application. So once different microservices are identified, how do you compose them to provide the application’s functionality?

This blog will discuss some of the recommended patterns on how to compose microservices together.

Aggregator Microservice Design Pattern

The first, and probably the most common, is the aggregator microservice design pattern.

In its simplest form, Aggregator would be a simple web page that invokes multiple services to achieve the functionality required by the application. Since each service (Service A, Service B, and Service C) is exposed using a lightweight REST mechanism, the web page can retrieve the data and process/display it accordingly. If some sort of processing is required, say applying business logic to the data received from individual services, then you may likely have a CDI bean that would transform the data so that it can be displayed by the web page.

Microservice Aggregator Design Pattern

Another option for Aggregator is where no display is required, and instead it is just a higher level composite microservice which can be consumed by other services. In this case, the aggregator would just collect the data from each of the individual microservice, apply business logic to it, and further publish it as a REST endpoint. This can then be consumed by other services that need it.

This design pattern follows the DRY principle. If there are multiple services that need to access Service A, B, and C, then its recommended to abstract that logic into a composite microservice and aggregate that logic into one service. An advantage of abstracting at this level is that the individual services, i.e. Service A, B, and C, and can evolve independently and the business need is still provided by the composite microservice.

Note that each individual microservice has its own (optional) caching and database. If Aggregator is a composite microservice, then it may have its own caching and database layer as well.

Aggregator can scale independently on X-axis and Z-axis as well. So if its a web page then you can spin up additional web servers, or if its a composite microservice using Java EE, then you can spin up additional WildFly instances to meet the growing needs.

Proxy Microservice Design Pattern

Proxy microservice design pattern is a variation of Aggregator. In this case, no aggregation needs to happen on the client but a different microservice may be invoked based upon the business need.

Microservice Proxy Design Pattern


Just like Aggregator, Proxy can scale independently on X-axis and Z-axis as well. You may like to do this where each individual service need not be exposed to the consumer and should instead go through an interface.

The proxy may be a dumb proxy in which case it just delegates the request to one of the services. Alternatively, it may be a smart proxy where some data transformation is applied before the response is served to the client. A good example of this would be where the presentation layer to different devices can be encapsulated in the smart proxy.

Chained Microservice Design Pattern

Chained microservice design pattern produce a single consolidated response to the request. In this case, the request from the client is received by Service A, which is then communicating with Service B, which in turn may be communicating with Service C. All the services are likely using a synchronous HTTP request/response messaging.

Microservice Chain Design Pattern

The key part to remember is that the client is blocked until the complete chain of request/response, i.e. Service <-> Service B and Service B <-> Service C, is completed. The request from Service B to Service C may look completely different as the request from Service A to Service B. Similarly, response from Service B to Service A may look completely different from Service C to Service B. And that’s the whole point anyway where different services are adding their business value.

Another important aspect to understand here is to not make the chain too long. This is important because the synchronous nature of the chain will appear like a long wait at the client side, especially if its a web page that is waiting for the response to be shown. There are workarounds to this blocking request/response and are discussed in a subsequent design pattern.

A chain with a single microservice is called singleton chain. This may allow the chain to be expanded at a later point.

Branch Microservice Design Pattern

Branch microservice design pattern extends Aggregator design pattern and allows simultaneous response processing from two, likely mutually exclusive, chains of microservices. This pattern can also be used to call different chains, or a single chain, based upon the business needs.

Microservice Branch Design Pattern

Service A, either a web page or a composite microservice, can invoke two different chains concurrently in which case this will resemble the Aggregator design pattern. Alternatively, Service A can invoke only one chain based upon the request received from the client.

This may be configured using routing of JAX-RS or Camel endpoints, and would need to be dynamically configurable.

Shared Data Microservice Design Pattern

One of the design principles of microservice is autonomy. That means the service is full-stack and has control of all the components – UI, middleware, persistence, transaction. This allows the service to be polyglot, and use the right tool for the right job. For example, if a NoSQL data store can be used if that is more appropriate instead of jamming that data in a SQL database.

However a typical problem, especially when refactoring from an existing monolithic application, is database normalization such that each microservice has the right amount of data – nothing less and nothing more. Even if only a SQL database is used in the monolithic application, denormalizing the database would lead to duplication of data, and possibly inconsistency. In a transition phase, some applications may benefit from a shared data microservice design pattern.

In this design pattern, some microservices, likely in a chain, may share caching and database stores. This would only make sense if there is a strong coupling between the two services. Some might consider this an anti-pattern but business needs might require in some cases to follow this. This would certainly be an anti-pattern for greenfield applications that are design based upon microservices.

Microservice Branch Shared Data Design Pattern

This could also be seen as a transition phase until the microservices are transitioned to be fully autonomous.

Asynchronous Messaging Microservice Design Pattern

While REST design pattern is quite prevalent, and well understood, but it has the limitation of being synchronous, and thus blocking. Asynchrony can be achieved but that is done in an application specific way. Some microservice architectures may elect to use message queues instead of REST request/response because of that.

Microservice Async Messaging Design Pattern

In this design pattern, Service A may call Service C synchronously which is then communicating with Service B and D asynchronously using a shared message queue. Service A -> Service C communication may be asynchronous, possibly using WebSockets, to achieve the desired scalability.

A combination of REST request/response and pub/sub messaging may be used to accomplish the business need.

Coupling vs Autonomy in Microservices is a good read on what kind of messaging patterns to choose for your microservices.

Hope you find these design patterns are useful.

What microservice design patterns are you using?

JavaOne Cloud, DevOps, Containers, Microservices etc. Track


Every year, for the past 19 years, JavaOne is the biggest and most awesome gathering of Java enthusiasts from around the world. JavaOne 2015 is the 20th edition of this wonderful conference. How many conferences can claim this? :)

Would you like to be part of JavaOne 2015? Sure, you can get notified when the registration opens and attend the conference. Why not take it a notch higher on this milestone anniversary?

Submit a session and become a speaker? Tips for Effective Sessions Submissions at Technology Conferences provide detailed tips on how to make the session title/abstract compelling for the program committee.

Have you been speaking at JavaOne for past several years? Don’t wait, and just submit your session today. The sooner you submit, higher the chances of program committee members voting on it. You know the drill!

Important Dates

  • Call for Papers closes April 29, 2015
  • Notifications for accepted and declined sessions: mid-June
  • Conference date: Oct 25 – 29

JavaOne Tracks

JavaOne conference is organized by tracks, and the tracks for this year are:

I’m excited and honored to co-lead the Java, DevOps, and the Cloud track with Bruno Borges (@brunoborges). The track abstract is:

The evolution of service-related enterprise Java standards has been underway for more than a decade, and in many ways the emergence of cloud computing was almost inevitable. Whether you call your current service-oriented development “cloud” or not, Java offers developers unique value in cloud-related environments such as software as a service (SaaS) and platform as a service (PaaS). The Java Virtual Machine is an ideal deployment environment for new microservice and container application architectures that deploy to cloud infrastructures. And as Java development in the cloud becomes more pervasive, enabling application portability can lead to greater cloud productivity. This track covers the important role Java plays in cloud development, as well as orchestration techniques used to effectively address the service lifecycle of cloud-based applications. Track sessions will cover topics such as SaaS, PaaS, DevOps, continuous delivery, containers, microservices, and other related concepts.

So what exactly are we looking for in this track?

  • How have you been using PaaS effectively for solving customer issues?
  • Why is SaaS critical to your business? Are you using IaaS, PaaS, SaaS all together for different parts of your business?
  • Have you used microservices in a JVM-based application? Lessons from the trenches?
  • Have you transformed your monolith to a microservice-based architecture?
  • How are containers helping you reduce impedance mismatch between dev, test, and prod environments?
  • Building a deployment pipeline using containers, or otherwise
  • Are PaaS and DevOps complimentary? Success stories?
  • Docker machine, compose, swarm recipes
  • Mesosphere, Kubernetes, Rocket, Juju, and other clustering frameworks
  • Have you evaluated different containers and adopted one? Pros and Cons?
  • Any successful practices around containers, microservices, and DevOps together?
  • Tools, methodologies, case studies, lessons learned in any of these, and other related areas
  • How are you moving legacy applications to the Cloud?
  • Are you using private clouds? Hybrid clouds? What are the pros/cons? Successful case studies, lessons learned.

These are only some of the suggested topics and are looking forward to your creative imagination. Remember, there are a variety of formats for submission:

  • 60 mins session or panel
  • Two-hour tutorial or hands-on lab
  • 45 mins BoFs
  • 5 mins Ignite talk

We think this is going to be the coolest track of the conference, with speakers eager to share everything about all the bleeding edge technologies and attendees equally eager to listen and learn from them. We’d like to challenge all of you to submit your best session, and make our job extremely hard!

Once again, make sure to read Tips for Effective Sessions Submissions at Technology Conferences for a powerful session submission. One key point to remember: NO vendor or product pitches. This is a technology conference!

Dilbert Technology Show

Links to Remember

JavaOne is where you have geekgasm multiple times during the day. This is going to be my 17th attendance in a row, and so looking forward to see you there!

Microservices, Monoliths, and NoOps

Monolithic Applications

A monolith application, in layman terms, is where entire functionality of the application is packaged together as a single unit or application. This unit could be JAR, WAR, EAR, or some other archive format, but its all integrated in a single unit. For example an online shopping website will typically consists of customer, product, catalog, checkout, and other features. Another example is of a movieplex. Such an application would typically consist of show booking, add/delete movie, ticket sales, accrue movie points, and other features. In case of a monolithic application, all these features are implemented and packaged together as one application.

Movieplex7 is one such canonical Java EE 7 sample application and the main features are shown below:

Movieplex7 Features

This application when packaged as a WAR would looks like:

Moviexplex WAR Package

The archive consists of some web pages that forms the UI. Classes implement the business logic, persistence, backing beans, etc. And finally there are some configuration files that define database connection, CDI configuration, etc.

More specifically, structure of the WAR looks like:

Movieplex7 WAR Structure

In this WAR structure, web pages are within the green box, all classes are within the orange box, and configuration files are within the blue box.

This application is somewhat modular as all the classes are neatly organized in package by different functionality. Web pages and configuration files follow a similar pattern as well.

Advantages of Monolithic Applications

There are a few advantages of this style of application:

  1. Well Known: This is how typically applications have been built so far. Its easy to conceptualize and all the code is in one place. Majority of existing tools, application servers, frameworks, scripts are able to deal with such kind of applications.
  2. IDE-friendly: Development environments, such as NetBeans, Eclipse, or IntelliJ, can be easily setup for such applications. IDEs are typically designed to develop, deploy, debug, profile a single application easily. Stepping through the code base is easy because the codebase is all together.
  3. Easy Sharing: A single archive, with all the functionality, can be shared between teams and across different stages of deployment pipeline.
  4. Simplified Testing: Once the application is deployed successfully, all the services, or features, are up and available. This simplifies testing as there are no additional dependencies to wait for in order for the testing to begin. Either the application is available, in which case all features are available, or the application is not available at all. Accessing or testing the application is simplified in either case.
  5. Easy Deployment: Easy to deploy since, typically, a single archive needs to be copied to one directory. The deployment times could vary but the process is pretty straight forward.

Disadvantages of Monolithic Applications

Monolith applications have served us well so far, and most likely will continue to work for some in the years to come. There are websites like Etsy that have 60 million monthly visitors and 1.5 billion monthly page views and are built/deployed as one large monolith. They have taken monoliths to an extreme where they are doing 50 deploys/day using a single large application. Unfortunately, most of the companies are not like that.

A monolithic application, no matter how modular, will eventually start to break down as the team grows, experienced developers leave and new ones join, application scope increases, new ways to access the applications are added, and so on. Take any monolith application that has spanned multiple years and teams and the entire code base will look like a bug ball of mud. That’s how software evolves especially when there is a pressure to deliver.

Lets look at some of the disadvantages of monolithic applications

  • Limited Agility: Every tiny change to the application means full redeployment of the archive. Consider the use case where only one piece of functionality in the application needs to be updated, such as booking or add/delete movie. This would require the entire application to be built again, and deployed again, even though other parts of the application has not changed. This means that developers will have to wait for the entire application to be deployed if they want to see the impact of quick change made in their workspace. Even if not intentional, but this may require tight coupling between different features of the application. This may not be possible all the time, especially if multiple developers are working on the application. This reduces agility of the team and the frequency by which new features can be delivered.
  • Obstacle for continuous delivery: The sample application used here is rather small so the time it takes to rebuild and deploy the archive would not be noticeable much. But a real-life application would be much bigger, and deployment times can be frustratingly long and slow. If a single change to the application would require entire application to be redeployed then this could become an obstacle to frequent deployments, and thus an impediment of continuous deployment. This could be a serious issue if you are serving a mobile application where users expect latest cool new feature all the time.
  • “Stuck” with Technology Stack: Choice of technology for such applications are evaluated and decided before the application development starts. Everybody in the team is required to use the same language, persistence stores, messaging system, and use similar tools to keep the team aligned. But this is like fitting a square peg in a round hole. Is MySQL an appropriate data store for storing graph databases? Is Java the most appropriate language for building front-end reactive applications? Its typically not possible to change technology stack mid stream without throwing away or rewriting significant part of existing application.
  • Technical Debt: “Not broken, don’t fix it” methodology is very common in software developed, more so for monolithic applications. This is convenient and enables to keep the application running.  A poor system design or badly written code is that much difficult to modify because other pieces of the application might be using it in unexpected ways. Software entropy of the system increases over a period of time unless it is refactored. Typically such an application is built over several years with the team that is maintaining the code base completely different from the one that created the application. This increases technical debt of the application and makes it that much harder to refactor the application later on.

What are Microservices?

The growing demand for agility, flexibility, and scalability to meet rapidly evolving business needs creates a strong need for a faster and more efficient delivery of software.

Meet Microservices!

Microservices is a software architectural style that require functional decomposition of an application. A monolithic application is broken down into multiple smaller services, each deployed in its own archive, and then composed as a single application using standard lightweight communication, such as REST over HTTP. The term “micro” in microservices is no indication of the LOCs in the service, it only indicates the scope is limited to a single functionality.

We’ve all been using microservices for a few years already. Think about a trivial mobile application can tell you the ratings of a hotel, find out the weather at your destination, book the hotel, locate directions to your hotel, find a nearby restaurant, and so on. This application is likely using different services such as Yelp, Google Maps, Yahoo Weather API, etc to accomplish these tasks. Each of this functionality is effectively running as an independent service and composed together in this single mobile application. Explosion of mobile apps, and their support for the growing business demand is also highlighted by Forrester’s four-tier engagement platform, and services are a key part of that.

Lets look at what are the characteristics of a microservice based application.

Characteristics of Microservices

Lets look at the characteristics of an application built using microservices.

  • Domain Driven Design: Functional decomposition of an application can be achieved using well-defined principles of Domain Driven Design by Eric Evans. This is not the only way to break down the applications but certainly a very common way. Each team is responsible for building the entire functionality around that domain or function of the business. Teams building a service include full range of developers, thus following the full-stack development methodology, and include skills for user interface, business logic, and persistence.
  • Single Responsibility Principle: Each service should have responsibility over a single part of the functionality, and it should do that well. This is one of the SOLID principles and has been very well demonstrated by Unix utilities.
  • Explicitly Published Interface: Each service publishes an explicitly defined interface and honors that all times. The consuming service only cares about that interface, and does not, rather should not, have any runtime dependency on the consumed service. The services agree upon the domain models, API, payload, or some other contract and they communicate using only that. A newer version of the interface may be introduced, but either the previous versions will continue to exist or the newer services are backwards compatible. You cannot break compatibility by changing contracts.
  • Independently Deploy, Upgrade, Scale, Replace: Each service can be independently deployed, and redeployed again, without impacting the overall system. This allows a service to be easily upgraded, for example to add more features. Each service can also scale independently on X-axis (horizontal duplication) or Z-axis (lookup oriented splits) as defined in Art of Scalability. Implementation of the service, or even the underlying technology stack, can change as long as the exact same contract is published. This is possible because other services rely only upon the published interface.
  • Potentially Heterogeneous/Polyglot: Implementation detail of the one service should not matter to another service. This enables the services to be decoupled from each other, and allows the team building the service to pick the language, persistence store, tools, methodology that is most appropriate for them. A service that requires to store data in a RDBMS can choose MySQL, and another service that needs to store documents can choose Mongo. Different teams can choose Java EE, NodeJS, Python, Vert.x, or whatever is most efficient for them.
  • Light-weight Communication: Services communicate with each other using a light-weight communication, such as REST over HTTP. This is inherently synchronous and so could have some potential bottle necks. An alternative mechanism is to use publish-subscribe mechanism that supports asynchronous messaging. Any of the messaging protocols such as AMQP, STOMP, MQTT, or WebSocket that meet the needs can be used there. Simple messaging implementations, such as ActiveMQ, that provide a reliable asynchronous fabric are quite appropriate for such usages. The choice of synchronous and asynchronous messaging is very specific to each service. They can even use a combination of the two approaches. Similarly the choice of protocol is very specific to each service but there is enough choice and independence for each team building the service.

Netflix is a poster child for microservices and several articles have been published on their adoption of microservices. A wide range of utilities that power their architecture are available at netflix.github.io.

Advantages of Microservices

  • Easier to develop, understand, and maintain: Code in a microservice is restricted to one function of the business and is thus easier to understand. IDEs can load the small code base very easily and keep the developers productive.
  • Starts faster than a monolith: Scope of each microservice is much smaller than a monolith and this leads to a smaller archive. As a result the deployment and the startup is much faster keeping developers productive.
  • Local change can be easily deployed: Each service can be deployed independent of other services. Any change local to the service can be easily made by the developer without requiring coordination with other teams. For example, performance of a service can be improved by changing the underlying implementation. As a result this keeps the agility of the microservice. This is also a great enabler of CI/CD.
  • Scale independently: Each service can scale independently using X-axis cloning and Z-axis partitioning based upon their need. This is very different from monolithic applications that may have very different requirements and yet must be deployed together.
  • Improves fault isolation: A misbehaving service, such as with a memory leak or unclosed database connections, will only affect that service as opposed to the entire monolithic application. This improves fault isolation and does not brings the entire application down, just a piece of it.
  • No long term commitment to any stack: Developers are free to pick language and stack that is best suited for their service. Even though the organizations may restrict the choice of technology but you are not penalized because of past decisions. It also enables to rewrite the service using better languages and technologies. This gives freedom of choice to pick a technology, tools, and frameworks.

Microservices may seem like a silver bullet that can solve significant amount of software problems. They serve a pretty good purpose but are certainly not easy. A significant operations overhead is required for these, and this article from Infoworld clearly points out.

with microservices, some technical debt is bound to shift from dev to ops, so you’d better have a crack devops team in place

This is very critical as now your one monolith is split across multiple microservices and they must talk to each other. Each microservice may be using a different platform, stack, persistent store and thus will have different monitoring and management requirements. Each service can then independently scale on X-axis and Z-axis. Each service can be redeployed multiple times during the day.

Microservices and NoOps

This imposes additional requirements on your infrastructure. These are commonly put together and called as NoOps. Essentially these are a set of services that provide a better process for deploying applications, and keep them running.

  • Service replication: Each service need to replicate, typically using X-axis cloning or Y-axis partitioning. Does each service need to build their logic to scale? For example, Kubernetes provide a great way to replicate services easily using Replication Controller.
  • Service discovery: Multiple services might be collaborating to provide an application’s functionality. This will require a service to discover other services. It could be tricky in a cloud environment where the services are ephemeral and possibly scale up and down. Resolving the services that are required for a service is thus a common functionality for all other services. Services need to register with a central registry and other services need to query this registry for resolving any dependencies. Netflix Eureka, Etcd, Zookeeper are some options in this space (more details).
  • Resiliency: Failure in software occurs, no matter how much and how hard you test it. The key question is not “how to avoid failure” but “how to deal with it”. This is all the more prominent in microservices where services are distributed all over the Internet. Its important for services to automatically take corrective action and ensure the user experience is not impacted. Michael Nygard’s book Release It! introducs Circuit Breaker pattern to deal with software resiliency. Netflix’s Hystrix provide an implementation of this design pattern (more details).
  • Service monitoring: One of the most important aspects of a distributed system is service monitoring and logging. This allows to take a proactive action, for example, if a service is consuming unexpected resources.

Refactoring into Microservices

Microservices also does not mean you need to throw away your existing application. Rather in majority (99.9%?) cases, you cannot throw away the application. So you’ve to build a methodology on how to refactor an existing application using microservices. However you need to bring your monolith to a stage where it is ready for refactoring. As Distributed big balls of mud highlight:

If you can’t built a monolith, what makes you think microservices are the answer?

Refactoring may not be trivial but in the long terms this has benefits which is also highlighted in the previously quoted article from Infoworld:

Refactoring a big monolithic application [using microservices] can be the equivalent of a balloon payment [for reducing technical debt] … you can pay down technical debt one service at a time

Functional decomposition of a monolith is very important otherwise it becomes a distributed monolith as opposed to a microservice based application.

Future Blogs

A subsequent blog on blog.arungupta.me will show how to refactor an existing Java EE application using microservices.

Some more questions that would be answered in subsequent blogs ….

  • How is it different from SOA?
  • Is REST the only way to exchange data? What messaging prorotocols?
  • Does microservices simplify/require CI/CD?
  • How is it related to Containers and DevOps? Are containers required to run microservices?
  • Are there any standards around microservices?
  • Are we pushing the problems around to orchestration?
  • What roles does PaaS play to enable microservices?
  • How can existing investment be leveraged?
  • Microservices Maturity Model