Search This Blog


Friday, April 17, 2015

Reactive Programming with Jersey and RxJava


I have the reactive programming fever right now and am having fun using RxJava (Reactive Extensions for Java).  When I found out about reactive support for Jersey, I figured a BLOG might be warranted.

Problem Domain

It's kinda nice to have problem domain. In our case company Acme had a monolithic retail web app application that sold products. There were many modules composing the monolith such as CheckOut, Account details, etc all contributing to a single deployable. One module among them was Product which had had core product information, pricing information and inventory that were obtained from a database.

The solution worked really well at first. However, over time, the company felt the following pain points -
  • One Deploy thus One Rollback - All or nothing
  • Long build time and testing
  • Information such as price, reviews etc were needed elsewhere as there was some logic associated with these thus having it centralized for consumption would be nice.
  • Coordination nightmare among teams working on different areas to make sure that all came together.
The Architects  decided to extract the concept of Product into separate services.  Over time, the company created separate web services for the base product, inventory, reviews and price. These were aligned with the different teams working on the respective areas and allowed for each area to evolve and scale independently.

All the web service calls were written using JAX-RS/Jersey. The company had totally adopted the micro services bandwagon.  A product would be created for the Web Site by calling the different services and aggregating the result as shown below  -

The following is a sample code that uses just one jersey client that demonstrates the above :
   public void getProduct(@PathParam("id") Long productId,
    @Suspended final AsyncResponse response) {
    try {
      // Get the Product
      BaseProduct baseProduct ="/products")

      // Get reviews of the Product
      Reviews reviews ="/productReviews")

      // Create a Result Product object - Base Product object does not have price and inventory
      Product resultProduct = resultProductFrom(baseProduct, reviews);

      // Obtain the Price for Result Product

      // Get Price and Inventory of Product Options and set the same
      for (Long optionId : optionId(resultProduct.getOptions())) {        
        Double price ="/productPrice")

        ProductInventory inventory =

        ProductOption option = resultProduct.getOption(optionId);


    catch (Exception e) {

ACME soon found that calling these services had a performance degradation compared to when they were a monolithic product service that obtained all product information using a database operation.The performance degradation was primarily attributed to the serial nature in which the requests were being invoked and results subsequently composed not counting the fact that a database join among disparate tables was far more performant. As an immediate fix, the direction from the Architects was to ensure these services were called in parallel when appropriate -
  • Call the Product Service to obtain core product data serially - You need base product information and its options
  • Call the Review Service asynchronously
  • Call the Pricing Service asynchronously to obtain the price of the Product and Options
  • Call the Inventory Service asynchronously to obtain the inventory of the different Product Options
The above worked and things were more performant, however, the code looked a mess due to the different CountDownLatches, composition logic and Futures they had in play. The Architects met again, hours were spent locked in a room while pizza's where delivered under their door until they heard of Reactive Programming in general and Reactive Jersey with RxJava in particular! Bling, Bling, bulbs coming to light illuminating their halo's and away they go reacting to their new idea. They found that Reactive Jersey promotes a clean API to handle parallel execution and composition while not having to worry about Countdown latches and the likes. The resulting code developed looked like -
  public void observableProduct(@PathParam("id") final Long productId,
    @Suspended final AsyncResponse response) {

    // An Observable of a Result Product from 
    Observable<Product> product =, reviews(productId),
      new Func2<BaseProduct, Reviews, Product>() {

        public Product call(BaseProduct product, Reviews review) {
          return resultProductFrom(product, review);

    // All Product
    Observable<Long> productIds = productAndOptionIds(product);
    // Observable of Options only
    Observable<Long> optionIds = productIds.filter(new Func1<Long, Boolean>() {

      public Boolean call(Long prodId) {
        return !prodId.equals(productId);

    // Set Inventory Data
        .zipWith(inventories(productId, optionIds).toList(), new Func2<Product, List<ProductInventory>, Product>() {

          public Product call(Product resultProduct, List<ProductInventory> productInventories) {
            for (ProductInventory inventory : productInventories) {
              if (!inventory.getProductId().equals(resultProduct.getProductId())) {
            return resultProduct;
        // Set Price Data
        .zipWith(prices(productIds).toList(), new Func2<Product, List<ProductPrice>, Product>() {

          public Product call(Product resultProduct, List<ProductPrice> prices) {
            for (ProductPrice price : prices) {
              if (price.getProductId().equals(resultProduct.getProductId())) {
              else {
            return resultProduct;
        }).observeOn( Action1<Product>() {

          public void call(Product productToSet) {

        }, new Action1<Throwable>() {

          public void call(Throwable t1) {
   * @return an Observable of the BaseProduct
  private Observable<BaseProduct> baseProduct(Long productId) {
    return RxObservable
   * @return An Observable of the Reviews
  private Observable<Reviews> reviews(Long productId) {
    return RxObservable
   * @return An Observable having Product and Option Ids
  private Observable<Long> productAndOptionIds(Observable<Product> product) {
    return product.flatMap(new Func1<Product, Observable<Long>>() {

      public Observable<Long> call(Product resultProduct) {
        return Observable.from(Iterables.concat(
          Lists.<Long> newArrayList(resultProduct.getProductId()),
          Iterables.transform(resultProduct.getOptions(), new Function<ProductOption, Long>() {

            public Long apply(ProductOption option) {
              return option.getProductId();
   * Inventories returns back inventories of the Primary product and options. 
   * However, for the primary product, no web service call is invoked as inventory of the main product is the sum of 
   * inventories of all options. However, a dummy ProductInventory is created to maintain order during final concatenation.
   * @param productId Id of the Product
   * @param optionIds Observable of OptionIds
   * @return An Observable of Product Inventory
  private Observable<ProductInventory> inventories(Long productId, Observable<Long> optionIds) {
    return Observable.just(new ProductInventory(productId, 0))
        .concatWith(optionIds.flatMap(new Func1<Long, Observable<ProductInventory>>() {

      public Observable<ProductInventory> call(Long optionId) {
        return RxObservable
            .resolveTemplate("productId", optionId).request().rx().get(ProductInventory.class);
   * @return An Observable of ProductPrice for product Ids
  private Observable<ProductPrice> prices(Observable<Long> productIds) {
    return productIds
        .flatMap(new Func1<Long, Observable<ProductPrice>>() {

          public Observable<ProductPrice> call(Long productId) {
            return RxObservable
                .resolveTemplate("productId", productId.toString()).request().rx()

Phew! A lot of code for something simple huh? The above code is demonstrated using jdk 7. With jdk 8 Lambda's, it should be far more succinct. However, I will admit that there is more code but not code that is not clear (hopefully). The important thing is that we are not dealing with the 'callback hell' associated with Java Futures or Invocation callbacks. That said, the performance difference between the serial execution and the Reactive Jersey version appears to strongly favor the Reactive Jersey version by a significant magnitude. By taking the serial call and turning in into parallel execution, I am certain you will achieve closer numbers to the Reactive Jersey version but at the cost of having to maintain latches etc.

Running the Example

An example application can be downloaded from here. The example does not have the Acme web site but has a Product-Gateway that serves as an orchestration layer. There are two resources in the Product-Gateway, one that returns the product using serial orchestration and a second that uses Reactive Jersey clients.  An Integration Test project exists where a client is used to invoke both the Serial and Reactive resources and logs an average response time.  Checkout the project, and simply execute a 'mvn install' at the root level of the project.


  • Reactive programming appears more verbose. Possible that I have not optimized it well enough, tips would be appreciated. jdk-8 Lamda's would regardless reduce the footprint.
  • Requires a mind shift regarding how you write code.
  • Aysnc could be introduced as required using Observable.subscribeOn() and Observable.observeOn() features as an when required. Read more about it on a Stack Overflow post.
  • It's a style of programming not a particular technology. RxJava seems to be emerging as the library of choice for Java.
  • Question whether something benefits from being Reactive before introducing Reactive code.
  • Jersey Reactive is a Glassfish implementation and not part of a JSR. So....
  • Monlith to Microservices might only be trading one problem for another so observe caution and be judicious in your selection of what you choose to make into a microservice
  • Batch request that fetches multiple items is not necessarily faster than multiple requests for a each item.
Finally, if there is a way to make my Reactive jersey code less verbose, let me know...all I ask is don't say collapse the services back to the monolith :-)

Monday, September 29, 2014

Spring @Value and javax.inject.Provider for Changeable Property Injection


With Spring and other DI containers one works with 'properties' that are used to configure values. Some values might needs to be changed dynamically at runtime while others might be statically configured. A changeable property lends itself to a javax.inject.Provider style where the property is obtained as required and used. Sure that one could inject a java.util.Properties class and use it to obtain the corresponding property. It just seems more in line with The Law of Demeter to inject the actual property or the next closest thing, i.e., a dedicated Provider of the property.

Spring has the support for injecting a property via @Value annotation as :
public class Foo {
 private Integer integer;
or Constructor Injection
public class Foo {

 private final Integer integer;
 public Foo(@Value("${fooVal}") Integer integer) {
   this.integer = integer;
or Setter Injection
public class Foo {

 private Integer integer;
 public setInteger(@Value("${fooVal}") Integer integer) {
   this.integer = integer;

Or if using JavaConfig:
public class FooConfig {
  public Foo foo(@Value("${fooVal}") Integer integer) {
    return new Foo(integer);

If fooVal never changes during the execution of the program, great, you inject it and then its set. What if it were to change, due to it being a configurable value? What would be nice to do is the following:
public class Foo {
  private final Provider<Integer> integerProvider;
  public Foo(Provider<Integer> integerProvider) {
    this.integerProvider = integerProvider;
  public void someOperation() {
    Integer integer = integerProvider.get(); // Current value of fooVal

public class FooConfig {
  public Foo foo(@Value("${fooVal}") Provider<Integer> integerProvider) {
    return new Foo(integer);
Looks pretty reasonable, however, when you attempt the same with Spring Java Config, you end up with:
org.springframework.beans.TypeMismatchException: Failed to convert value of type 'java.lang.String' to required type 'java.lang.Integer'; nested exception is java.lang.IllegalArgumentException: MethodParameter argument must have its nestingLevel set to 1
 at org.springframework.beans.TypeConverterSupport.doConvert(
 at org.springframework.beans.TypeConverterSupport.convertIfNecessary(

This should just work. Simplest way I could get it to work is to fool spring that the nesting level of the Provider<Integer&gt; is actual one level lesser. The way I did it was to create my own org.spring.framework.beans.TypeConverter that delegated to the default TypeConverter used by Spring, i.e., the SimpleTypeConverter:
public class CustomTypeConverter implements TypeConverter {
  private final SimpleTypeConverter simpleTypeConverter;

  public CustomTypeConverter() {
    simpleTypeConverter = new SimpleTypeConverter(); // This is the default used by Spring

  public <T> T convertIfNecessary(Object newValue, Class<T> requiredType,
    MethodParameter methodParam) throws IllegalArgumentException {
    Type type = methodParam.getGenericParameterType();

    MethodParameter parameterTarget = null;

    if (type instanceof ParameterizedType) {
      ParameterizedType paramType = ParameterizedType.class.cast(type);
      Type rawType = paramType.getRawType();

      if (rawType.equals(Provider.class)
          && methodParam.hasParameterAnnotation(Value.class)) { 
        // If the Raw type is javax.inject.Provider, reduce the nesting level
        parameterTarget = new MethodParameter(methodParam); 
       // Send a new Method Parameter down stream, don't want to fiddle with original

    return simpleTypeConverter.convertIfNecessary(newValue, requiredType, parameterTarget);
  ...// Delegate other methods to simpleTypeConverter
With the above you could do the following:
public class SpringTest {

  public void test() {
    AnnotationConfigApplicationContext context = new AnnotationConfigApplicationContext();
    context.getBeanFactory().setTypeConverter(new CustomTypeConverter());
  public static class SimpleConfig {
    @Bean(name = "props") // Change this to a something that detects file system changes and pulls it in
    public PropertyPlaceholderConfigurer propertyPlaceHolderConfigurer() {
      PropertyPlaceholderConfigurer configurer = new PropertyPlaceholderConfigurer();
      Resource location = new ClassPathResource("");
      return configurer;
    public SimpleBean simpleBean(@Value("${fooVal}") Provider<Integer> propValue,
      @Value("${barVal}") Provider<Boolean> booleanVal) {
      return new SimpleBean(propValue, booleanVal);
  public static class SimpleBean {
    private final Provider<Integer> propVal;
    private final Provider<Boolean> booleanVal;
    public SimpleBean(Provider<Integer>, Provider<Boolean> booleanVal) {
      this.propVal = propValue;
      this.booleanVal = booleanVal;
    public void print() {
      System.out.println(propVal.get() + "," + booleanVal.get());
Now with the above if your properties changes, as you are injecting a javax.inject.Provider to the SimpleBean, at runtime, it will obtain the current value to use.

All this is great, I understand that I am hacking Spring to get what I want, is this the best way to handle something like dynamically changeable properties? Thoughts and suggestion welcomed.

Update :
I filed an enhancement request with Spring maintainers and they were fast to respond with this actually being a bug and are back porting fixes to previous versions as well as ensuring current and future versions have this feature available. 

Sunday, September 14, 2014

Dynamic Service Discovery with Apache Curator

Service Discovery is a core part of most Service Oriented Architectures in particular and Distributed Network Component Architectures in general.

Traditional usages to find the location of a service have been to use Local file system, Shared File System, Database or some 3rd party Service Registry that statically define service locations. In many cases, one knows exactly the number of services that will be used and their deployment locations thus leading to a statically defined registry as an ideal candidate for discovery. In larger scale systems statically defined locations can become harder to manage as -
  • Number of services in the ecosystem grow (micro-services)
  • Number of deployment 'environments' grow - Test Environment 1, Test Enviroment 2...etc
  • Elasticity scalability becomes desirable - More important with Cloud based environments
  • Multiple services versions need to be supported simultaneously (A/B test)
Enter the concept of Dynamic Service Registration and Discovery. Services on becoming available register themselves as a member ready to participate with some 'registry'. Service Consumers are notified of the availability of the new service and avail its usage. If the service terminates for whatever reason, the Service Consumer is notified of the same and can stop availing its usage. This is not a novel concept and has likely been addressed in different ways.

Any Dynamic Service Registration System presents the following challenges -
  • Availability - The Service Registry can become the SPOF (Single Point Of Failure) for an entire ecosystem
  • Reliability - Keeping the registry and thus the service consumers in sync with which service instance is currently available. A service might go down and if the registry and service consumer are not notified immediately, service calls might fail.
  • Generality - Registry that supports multiple operating systems and technologies.
Again, the above is solvable by different ways. The benefit we have is that there are a number of tools that intrinsically support the requirements of a dynamic registry such as ZooKeeper, Netflix Eureka, Linked In's Dynamic Discovery, Doozer etc.  Jason Wilder has an excellent BLOG on Open Source Service Discovery and a comparison of popular open source service registries.

I have been looking into ZooKeeper and thus will share a few of my findings on using it but before I do the same, I would like to take a moment to discuss Load Balancing.

In standard H/A architectures a Load Balancer serves to balance traffic across the different service instances as shown below -

The same works well but poses the following challenges -
  • A new Load Balancer for every service and every environment
  • An additional hop between a Service Consumer and Service
  • Adding/Removing services per need and time to do the same
  • Intelligent Routing and Supporting Service Degradation
A Dynamic Service Registry could help with the mentioned concerns  where different instances of Service C register themselves with ZooKeeper and Service A and Service B on obtaining locations of Service C execute calls directly to them  -

ZooKeeper's Ephermal nodes serve well to maintain registered service instances as an Ephermal node only exists as long the client is alive, so if a client dies, the Ephermal node is deleted.  It is possible to use ZooKeeper directly and develop code that will register services, discover registered service etc. However, there is a library called Apache Curator that was initially created by Netflix to make writing ZooKeeper based application easier and more reliable. It was then incorporated as an Apache Project. Curator provides 'recipes' to common use cases one would use ZooKeeper for, i.e., Leader Election, Queue, Semaphore, Distributed Locks etc. In addition to these 'recipes', it also provides a mechanism for Service Registration and Discovery.

The remainder of this BLOG will look at providing a simple example that demonstrates the use of Apache Curator in the context of a simple JAX-RS web service that gets registered and is subsequently dynamically discovered by clients. For the sake of the example, I will continue to use the Notes JAX-RS Web Service that I have used on previous posts.

Service Registration -

In the example domain, all services will be registered under the ZooKeeper path "/services".  When the Notes Service starts, it registers itself with ZooKeeper as an ephermal node in the path "/services/notes/XXXXXXXX", where XXX indicates the ephermal instance.

A simple class called ServiceRegistrar is responsible for registering the Notes Service:
public class ServiceRegistrarImpl implements ServiceRegistrar {
  private static final String BASE_PATH = "/services";
  private static final String SERVICE_NAME = "notes";  

  private final CuratorFramework client; // A Curator Client
  private ServiceDiscovery<InstanceDetails> serviceDiscovery;
  // This instance of the service
  private ServiceInstance<InstanceDetails> thisInstance;
  private final JsonInstanceSerializer<InstanceDetails> serializer; 
  public ServiceRegistrarImpl(CuratorFramework client, int servicePort) throws UnknownHostException, Exception {
    this.client = client;
    serializer = new JsonInstanceSerializer<>(InstanceDetails.class); // Payload Serializer
    UriSpec uriSpec = new UriSpec("{scheme}://{address}:{port}"); // Scheme, address and port
    thisInstance = ServiceInstance.<InstanceDetails>builder().name(SERVICE_NAME)
      .address(InetAddress.getLocalHost().getHostAddress()) // Service information 
      .payload(new InstanceDetails()).port(servicePort) // Port and payload
      .build(); // this instance definition

  public void registerService() throws Exception {
    serviceDiscovery = ServiceDiscoveryBuilder.builder(InstanceDetails.class)
    serviceDiscovery.start(); // Registers this service
  public void close() throws IOException {
    try {
    catch (Exception e) {"Error Closing Discovery", e);
A Servlet Listener is defined whose responsibility is to start the registrar -
public class ServiceRegistrarListener implements ServletContextListener {
  private static final Logger LOG = Logger.getLogger(ServiceRegistrarListener.class);
  public void contextDestroyed(ServletContextEvent sc) {  

  public void contextInitialized(ServletContextEvent sc) {
    try {
    catch (Exception e) {
      LOG.error("Error Registering Service", e);
      throw new RuntimeException("Exception Registering Service", e);

Service Discovery -
A Simple Discoverer class is shown below which uses the Curator library in obtaining instances of registered Notes Services. Curator provides a ServiceDiscovery class that is made aware of a Service Registry. A ServiceProvider is then used to obtain registered service instances for a particular service.
public class ServiceDiscoverer {
  private static final Logger LOG = Logger.getLogger(ServiceDiscovery.class);
  private static final String BASE_PATH = "/services";

  private final CuratorFramework curatorClient;

  private final ServiceDiscovery<InstanceDetails> serviceDiscovery;

  private final ServiceProvider<InstanceDetails> serviceProvider;

  private final JsonInstanceSerializer<InstanceDetails> serializer;

  private final List<Closeable> closeAbles = Lists.newArrayList();

  public ServiceDiscoverer(String zookeeperAddress, String serviceName) throws Exception {
    curatorClient = CuratorFrameworkFactory.newClient(zookeeperAddress,
      new ExponentialBackoffRetry(1000, 3)); // Ideally this would be injected

    serializer = new JsonInstanceSerializer<>(InstanceDetails.class); // Payload Serializer

    serviceDiscovery = ServiceDiscoveryBuilder.builder(InstanceDetails.class).client(curatorClient)
        .basePath(BASE_PATH).serializer(serializer).build(); // Service Discovery

    serviceProvider = serviceDiscovery.serviceProviderBuilder().serviceName(serviceName).build(); // Service Provider for a particular service

  public void start() {
    try {
      closeAbles.add(0, serviceDiscovery);
      closeAbles.add(0, serviceProvider);
    catch (Exception e) {
      throw new RuntimeException("Error starting Service Discoverer", e);

  public void close() {
    for (Closeable closeable : closeAbles) {
      // Close all
  public ServiceInstance<InstanceDetails> getServiceUrl() {
    try {
      return instance = serviceProvider.getInstance();
    catch (Exception e) {
      throw new RuntimeException("Error obtaining service url", e);
The ServiceProvider is used in the Notes Service Client as follows:
public class NotesClientImpl implements NotesClient {
  private static final Logger LOG = Logger.getLogger(NotesClientImpl.class);

  public static final String SERVICE_NAME = "notes";

  private final Client webServiceClient; // JAX-RS Client

  private final ServiceDiscoverer serviceDiscoverer; // Service Disoverer

   * @param getNotesServerUrl() Server URI
   * @throws Exception
  public NotesClientImpl(String zookeeperAddress) throws Exception {
    serviceDiscoverer = new ServiceDiscoverer(zookeeperAddress, SERVICE_NAME);

    ClientConfig config = new ClientConfig();
    webServiceClient = ClientBuilder.newClient(config);

  public NoteResult create(Note note) {
    ServiceInstance<InstanceDetails> instance = serviceDiscoverer.getServiceInstance();
    try {
      return webServiceClient
          .post(Entity.entity(note, MediaType.APPLICATION_XML_TYPE), NoteResult.class);
    catch (ProcessingException e) {
      // If a ProcessingException occurs, the current ServiceInstance is marked as down
      throw e;
In the above code, a ServiceInstance is obtained from the ServiceDiscoverer. It is important to 'not' locally hold onto the instance but use it during the processing of a call and then let go of it. If a occurs say due to a Service Instance dying abruptly during processing for the request, then, notifying Curator to not use that instance for a certain period of time is the recommended direction. There is a time out delay between when the service instance dies and ZooKeeper times out the ephermal node associated with the service. So unless the instance is marked as being in an error state on the service client, the ServiceDiscoverer would continue to provide the downed service instance for usage.

Integration Test -

A simple Integration Test has been shown where a number of Notes Server instances are started up and they register themselves with a test ZooKeeper instance. The Notes Client is able to round robin to each of those instances. Each of the servers are designed to randomly die after a period of time leading to them being un-registered .When a Notes Server Instance goes down, it will no longer be invoked as part of the test. There is a Notes Server started that runs for a long time and acts as a fall back for demonstration completeness.
public class SimpleIntegrationTest {
  private TestingServer testServer; // ZooKeeper Server

  private NotesServer longRunningNotesServer; // A Notes Server that does not die for the demo
  private static final int MAX_SERVERS = 10; // Max Servers to use
  public void setUp() throws Exception {
    testServer = new TestingServer(9001);   // Start the Curator ZooKeeper Test Server

  public void tearDown() throws IOException {...}

  public void integration() throws Exception {
    CyclicBarrier barrier = new CyclicBarrier(MAX_SERVERS + 1);
    for (Integer port : ports()) { // Start Instances of Notes Server on different ports
      new NotesServer(port, barrier).start(); // Start an Instance of the Notes Server at the provided port
    longRunningNotesServer = new NotesServer(9210, 500000L, null); // Start a Long Running Notes Server
    barrier.await(); // Wait for all Servers start
    NotesClient notesClient = new NotesClientImpl(testServer.getConnectString());
    List<NoteResult> noteResults = Lists.newArrayList();
    int i = 0;
    int maxNotes = 10000;
    while (i < maxNotes) {
      try {
        Note note = new Note("" + i, "" + i);
        NoteResult result = notesClient.create(note);
      catch (Exception e) {
        e.printStackTrace(); // If an error occurs on one instance, try again

    // Ensure that all Notes Servers have been utilized
    assertEquals(MAX_SERVERS +1, notesClient.getServiceInstancesUsed().size());
  private static Set<Integer> ports() {
    // Provide a unique set of ports for running multiple Notes Server
The example itself can be downloaded from HERE. Execute the project using:
>mvn install

Thoughts -
  • ZooKeeper is CP so in the event of Partition, writes can suffer.
  • ZooKeeper is SPOF of the ecosystem. Should have a contingency strategy.
    • Linked In's Dynamic Discovery Architecture addresses something like this but has a lot of moving parts
    • Simplicity of a Load Balancer makes one think YAGNI
      • Tools like Puppet Labs might make a lesser case for burden of standing up environments as well
  • Service Discovery allow clients to be 'smart'. Support for Service Degradation, alerting, and monitoring kind of go hand in hand.
  • Service Registration and De-Registration can be smartly built as well. For example, a Service could itself determine it's not healthy for whatever reason and de-registerer itself from the registry.
  • Supporting multiple environments where an ecosystem is self discoverable is simpler than having to provide overrides across different environments.
  • Managing  downed nodes with Curator needs to be handled carefully. Idempotence is your friend here.
  • Curator can definitely help with intelligent load balancing and different Load Balancing Strategies
    • Round Robin
    • Random
    • Custom one that involves Service Degradation
  • Curator provides a REST service that helps non-java clients wishing to participate in service registration and discovery.
  • If Netflix, Linked In and Amazon do something like this, it warrants looking into :-)

Friday, August 22, 2014

JVM Native memory leak

The Problem

There are times I learn something new of something that has been known since ages. Today was one of those times. Had a  Java Virtual Machine (JVM) whose heap usage looked great but the JVM process ended up growing native memory and finally died with an exception stating unable to allocate native memory. One such example might look like:
java.lang.OutOfMemoryError at Method) at at 

# There is insufficient memory for the Java Runtime Environment to continue.
# Native memory allocation (malloc) failed to allocate 184 bytes for AllocateHeap
# Possible reasons:
#   The system is out of physical RAM or swap space
#   In 32 bit mode, the process size limit was hit
# Possible solutions:
#   Reduce memory load on the system
#   Increase physical memory or swap space
#   Check if swap backing store is full
#   Use 64 bit Java on a 64 bit OS
#   Decrease Java heap size (-Xmx/-Xms)
#   Decrease number of Java threads
#   Decrease Java thread stack sizes (-Xss)
#   Set larger code cache with -XX:ReservedCodeCacheSize=
# This output file may be truncated or incomplete.

The application in question that manifested the above is a simple jersey web service that serves up and consumes JSON using JSON JAXB (Clue...). Initial interrogation yielded the following:
  • Thread usage is normal (-Xss settings has no benefit)
  •  –XXMaxDirectMemorySize - No difference
  • Heap usage is healthy
  • Heap dumps look healthy - no leaks
  • Perm Gen is looking good
  • Native file descriptors are within normal usage limits
  • creation did not appear to cause issues
  • JDK version latest (1.7.X)
  • Happens on Virtual machines or stand alone boxes
Some Stack Overflow and other assistance:
  • Using the G1 Garbage collector
  • Or Setting Xms and Xmx to smaller values, i.e., lowering heap size
Right, so using G1 Garbage collector worked, or setting max heap to a lower value also worked? WTF!

What happened?

Creation of  a JAXB Context on every request/response. The application in question was not caching JAXB Contexts but was creating them on every request/response. Initial thought is, yeah, that is a performance problem, but why will it count towards native process memory growth? Consider the following benign code:

public class SimpleTest extends Thread { 
  public void run() {
    while (!Thread.currentThread().isInterrupted()) {
      JSONJAXBContext context = new JSONJAXBContext(SomeDTOClass.class); // From jersey-json.jar

  public static void main(String args[]) {
    new SimpleTest().start();

Well, the problem is that JAXB Contexts indirectly involve in the invoking of native memory allocations. How does it do it and what happens underneath? Well, that is an exercise for the reader to figure out and comment on this BLOG with findings ;-).

The root cause can be dialed down to the following issue posted:

public class Bug {
  public static void main( String args[] ) {
    while ( true ) {
      /* If ANY of these two lines is not commented, the JVM runs out of memory */
      final Deflater deflater = new Deflater( 9, true );
      final Inflater inflater = new Inflater( true );

The same code that might work without native OOM:
public class Bug {
  public static void main( String args[] ) {
    int count = 0;
    while ( true ) {
      final Deflater deflater = new Deflater( 9, true );
      final Inflater inflater = new Inflater( true );
      if (i % 100 == 0) {
       System.gc(); // Don't, please don't do this!

As shown above, you might allocate 'small' objects in the JVM but underneath they might end up allocating native memory for their operations. The intention is noble, that when the objects are finalized, the native operating system memory will be de-allocated as well. However, the freeing of the native memory is contingent on the Garbage collector running and leading to finalization of these 'small' objects. Its a matter of relative pace, the native memory is growing far faster compared to JVM heap and the lack of Garbage collection occurring is what leads to running out of native space. When using the G1 collector or lowering max heap settings, garbage collection occurs more often, thus preventing the native heap from exploding over time.


Sunday, April 20, 2014

jersey/jaxrs 2.X example using spring JavaConfig and spring security


I had blogged previously on using jersey/JAX-RS 2.0 when it was in a pre-release. Since then Jersey/JAX-RS 2.0 has released and undergone a few versions. As I have recently been working with the jersey 2.7-spring integration and spring security and figured I'd share an example.


As is quite standard with me, I like to 'manufacture' requirements. The application being developed is a Notes application that allows one to perform CRUD operations for simple notes. So, for the Web Service being created, I have the following requirements:
  1. No XML- You get type safety during refactoring
  2. Spring Dependency Injection with Java Config. The official jersey/spring3 example is very nice but does not demonstrate Java Config usage.
  3. Jersey Should manage Resource classes an have Spring service objects injected into them
  4. Security should be enabled via Spring Security. Only users with the ROLE 'notesuser' should be able to create a note.

The Project

Look Ma, no XML

This example will use Servlet 3.X specification, so we eliminate web.xml. There are many ways one can eliminate XML using Servlet 3.X, but this example focuses on the usage of javax.servlet.ServletContainerInitializer. With a ServletContainerInitializer one can programatically register Servlets, Filters and Listeners thus giving you type safety that can be useful during refactoring. As we are in Spring Land, rather than utilize ServletContainerInitializer, an extension of Spring's WebApplicationInitializer is used as shown below:
public class NotesApplicationInitializer implements WebApplicationInitializer {

  private static final String SPRING_SECURITY_FILTER_NAME = "springSecurityFilterChain";
  @Override public void onStartup(ServletContext ctx) throws ServletException {
    // Listeners

    ctx.setInitParameter(ContextLoader.CONFIG_LOCATION_PARAM, SpringConfig.class.getName());

    // Register Spring Security Filter chain
    ctx.addFilter(SPRING_SECURITY_FILTER_NAME, DelegatingFilterProxy.class)
          EnumSet.<DispatcherType> of(DispatcherType.REQUEST, DispatcherType.FORWARD), false, "/*");

    // Register Jersey 2.0 servlet
    ServletRegistration.Dynamic servletRegistration = ctx.addServlet("Notes",
    servletRegistration.setInitParameter("", NotesApplication.class.getName());
The WebApplicationInitializer does the following:
  1. Sets up a Spring ContextLoaderListener
  2. Tells the Context loader the location of the Spring Java Config that is to be used for Services
  3. Registers the Spring Security Filter chain
  4. Registers the Jersey Servlet Container by providing it the NotesApplication as the class to use for Resource and Provider management
Of note is the fact that the NotesApplicationInitializer has been set to highest precedence as that will ensure that it is executed before any other WebApplicationInitializer provided by accompanying jars. If for example, the SpringWebApplicationInitializer from jerseyspring3.jar gets loaded, then it attempts to find a spring  applicationContext.xml and will fail as our example does not use spring xml style bean definitions.

Spring Dependency Injection

In conjunction with the registration of the Spring Java Config in the WebApplicationInitializer shown above, the Java Config enables the Services and Spring Security filter chain via the following:
public class SpringConfig {
I have not delved into the details of the @Enable annotations but they are in line with Spring's style to import dependencies. For the scope of this example, you can assume that @EnableNotesSecurity results in the importing of the notes Java services and @EnableNotesSecurity imports the security configuration.

Jersey Should Manage the Resource Classes

I wanted all my JAX-RS resources managed by Jersey and not Spring. I did not want to annotate my JAX-RS classes with @Component + Classpath-scanning and/or have them defined in a Java Config which would then result in Spring managing them. All the Resources and relevant providers are registered with Jersey by extending the class as shown below:
public class NotesApplication extends Application {
  public Set<Class<?>> getClasses() {
    return ImmutableSet.<Class<?>>of(NotesResource.class, 
      NoteNotFoundExceptionMapper.class, LoggingFilter.class, AccessDeniedExeptionMapper.class);
Note that the 'service' java classes, such as (managed by Spring) are made available via dependency injection into the NotesResource via Jersey's Spring-HK2 bridge.

Security Should be enabled Via Spring Security

For this example,I have very simple Spring Security filter chain set up that ensures a POST to create a Note can only be done by a user who has the ROLE 'notesuser' provided in a HTTP header while making the call. It is trivial, but hey, this is an example :-) The goal was to demonstrate the use of a @PreAuthorize annotation on the create a Note and how it can be made to work when Jersey manages the resource and not spring. The Notes resource create method looks like the following:
@Produces({ MediaType.APPLICATION_XML })
@Consumes({ MediaType.APPLICATION_XML })
public class NotesResource {
  // Jersey object injection
  private final UriInfo uriInfo;
  // Spring object injection
  private final NotesService notesService;
  // Note that UriInfo is obtained from Jersey but the NotesService is a spring dependency
  public NotesResource(@Context UriInfo uriInfo, NotesService notesService) {
    this.uriInfo = uriInfo;
    this.notesService = notesService;
  @PreAuthorize("hasRole('notesuser')") // Pre Authorize role that allows only notesuser to create the note
  public Response create(Note note) {
    LOG.debug("Creating a note:" + note);
    NoteResult result = createNote(note);
    return Response.created(result.getLocation()).entity(result).build();
As the NotesResource is managed by Jersey, the only way I could figure to enforce the @PreAuthorize annotation was via AspectJ weaving of the resource class. The same is accomplished by adding the aspectJ maven plugin which is used during the build process to enhance the JAX-RS Resource classes:

Running the Example

The example is provided as a maven project. I am utilizing the concept of a maven 'bom' for the first time for managing the related jersey dependencies. Execute a 'mvn install' to see the example in action or execute a 'mvn jetty:run' on the web project to start the service. The example itself can be downloaded from here.


The client will send the 'role' as a header parameter as shown below for creating a note:
  public NoteResult create(Note note, String role) {
        .header("role", role)
        .post(Entity.entity(note, MediaType.APPLICATION_XML_TYPE), NoteResult.class);

Integration Test

Integration test starts up a web container and executes the client operations. The following shows the spring-security role being honored:
  @Test(expected = ForbiddenException.class)
  public void roleAbsent() {
    Note note = new Note();
    note.setContent("Something to say");
    client.create(note, "Fake Role"); // Should throw a JAX-RS ForbiddenException
    fail("Should not have created a note as the role provided is not supported");
Running the equivalent of the above with the role of 'notesuser' will allow the creation of the note. The life-cycle integration test in the attached example demonstrates the same succeeding. The example should be devoid of web.xml. If anyone coming across this blog has integrated Jersey/Spring/Spring-security in a different way, please share, would love to hear. That's all folks! Enjoy!