Certain methods exposed in Map have certain performance consequences when used with Data Grid, such as size() , values() , keySet() and entrySet() . Specific methods on the keySet, values and entrySet are fine for use please see their Javadoc for further details.

Attempting to perform these operations globally would have large performance impact as well as become a scalability bottleneck. As such, these methods should only be used for informational or debugging purposes only.

It should be noted that using certain flags with the withFlags() method can mitigate some of these concerns, please check each method’s documentation for more details.

Further to simply storing entries, Data Grid’s cache API allows you to attach mortality information to data. For example, simply using put(key, value) would create an immortal entry, i.e., an entry that lives in the cache forever, until it is removed (or evicted from memory to prevent running out of memory). If, however, you put data in the cache using put(key, value, lifespan, timeunit) , this creates a mortal entry, i.e., an entry that has a fixed lifespan and expires after that lifespan.

In addition to lifespan , Data Grid also supports maxIdle as an additional metric with which to determine expiration. Any combination of lifespans or maxIdles can be used.

Data Grid’s Cache class contains a different 'put' operation called putForExternalRead . This operation is particularly useful when Data Grid is used as a temporary cache for data that is persisted elsewhere. Under heavy read scenarios, contention in the cache should not delay the real transactions at hand, since caching should just be an optimization and not something that gets in the way.

To achieve this, putForExternalRead() acts as a put call that only operates if the key is not present in the cache, and fails fast and silently if another thread is trying to store the same key at the same time. In this particular scenario, caching data is a way to optimise the system and it’s not desirable that a failure in caching affects the on-going transaction, hence why failure is handled differently. putForExternalRead() is considered to be a fast operation because regardless of whether it’s successful or not, it doesn’t wait for any locks, and so returns to the caller promptly.

To understand how to use this operation, let’s look at basic example. Imagine a cache of Person instances, each keyed by a PersonId , whose data originates in a separate data store. The following code shows the most common pattern of using putForExternalRead within the context of this example:

// Id of the person to look up, provided by the application
PersonId id = ...;

// Get a reference to the cache where person instances will be stored
Cache<PersonId, Person> cache = ...;

// First, check whether the cache contains the person instance
// associated with with the given id
Person cachedPerson = cache.get(id);

if (cachedPerson == null) {
   // The person is not cached yet, so query the data store with the id
   Person person = dataStore.lookup(id);

   // Cache the person along with the id so that future requests can
   // retrieve it from memory rather than going to the data store
   cache.putForExternalRead(id, person);
} else {
   // The person was found in the cache, so return it to the application
   return cachedPerson;
}

Note that putForExternalRead should never be used as a mechanism to update the cache with a new Person instance originating from application execution (i.e. from a transaction that modifies a Person’s address). When updating cached values, please use the standard put operation, otherwise the possibility of caching corrupt data is likely.

Flags are applied to regular cache methods to alter the behavior of certain methods. For a list of all available flags, and their effects, see the Flag enumeration. Flags are applied using AdvancedCache.withFlags() . This builder method can be used to apply any number of flags to a cache invocation, for example:

advancedCache.withFlags(Flag.CACHE_MODE_LOCAL, Flag.SKIP_LOCKING)
   .withFlags(Flag.FORCE_SYNCHRONOUS)
   .put("hello", "world");

Cache-level events occur on a per-cache basis, and by default are only raised on nodes where the events occur. Note in a distributed cache these events are only raised on the owners of data being affected. Examples of cache-level events are entries being added, removed, modified, etc. These events trigger notifications to listeners registered to a specific cache.

Please see the Javadocs on the org.infinispan.notifications.cachelistener.annotation package for a comprehensive list of all cache-level notifications, and their respective method-level annotations.

@Listener (clustered = true)
public class MyClusterListener { .... }

public class SpecificKeyFilter implements KeyFilter<String> {
    private final String keyToAccept;

    public SpecificKeyFilter(String keyToAccept) {
      if (keyToAccept == null) {
        throw new NullPointerException();
      }
      this.keyToAccept = keyToAccept;
    }

    public boolean accept(String key) {
      return keyToAccept.equals(key);
    }
}

...
cache.addListener(listener, new SpecificKeyFilter("Only Me"));
...

@Listener
public class MyRetryListener {
  @CacheEntryModified
  public void entryModified(CacheEntryModifiedEvent event) {
    if (event.isCommandRetried()) {
      // Do something
    }
  }
}

Cache manager-level events occur on a cache manager. These too are global and cluster-wide, but involve events that affect all caches created by a single cache manager. Examples of cache manager-level events are nodes joining or leaving a cluster, or caches starting or stopping.

See the org.infinispan.notifications.cachemanagerlistener.annotation package for a comprehensive list of all cache manager-level notifications, and their respective method-level annotations.

By default, all async notifications are dispatched in the notification thread pool. Sync notifications will delay the operation from continuing until the listener method completes or the CompletionStage completes (the former causing the thread to block). Alternatively, you could annotate your listener as asynchronous in which case the operation will continue immediately, while the notification is completed asynchronously on the notification thread pool. To do this, simply annotate your listener such:

Asynchronous Listener

@Listener (sync = false)
public class MyAsyncListener {
   @CacheEntryCreated
   void listen(CacheEntryCreatedEvent event) { }
}

Blocking Synchronous Listener

@Listener
public class MySyncListener {
   @CacheEntryCreated
   void listen(CacheEntryCreatedEvent event) { }
}

Non-Blocking Listener

@Listener
public class MyNonBlockingListener {
   @CacheEntryCreated
   CompletionStage<Void> listen(CacheEntryCreatedEvent event) { }
}

Non-blocking APIs are powerful in that they provide all of the guarantees of synchronous communications - with the ability to handle communication failures and exceptions - with the ease of not having to block until a call completes.  This allows you to better harness parallelism in your system.  For example:

Set<CompletableFuture<?>> futures = new HashSet<>();
futures.add(cache.putAsync(key1, value1)); // does not block
futures.add(cache.putAsync(key2, value2)); // does not block
futures.add(cache.putAsync(key3, value3)); // does not block

// the remote calls for the 3 puts will effectively be executed
// in parallel, particularly useful if running in distributed mode
// and the 3 keys would typically be pushed to 3 different nodes
// in the cluster

// check that the puts completed successfully
for (CompletableFuture<?> f: futures) f.get();

There are 4 things in Data Grid that can be considered to be on the critical path of a typical write operation. These are, in order of cost:

Using the async methods will take the network calls and marshalling off the critical path.  For various technical reasons, writing to a cache store and acquiring locks, however, still happens in the caller’s thread.

Data Grid Server disables the Memcached endpoint by default. If you enable the Memcached endpoint, you should configure caches with a suitable encoding for Memcached clients.

<memcached-connector cache="memcachedCache" client-encoding="application/json">

Encoding allows dealing with a certain data format during API calls (map, listeners, stream, etc) while the format effectively stored is different.

The data conversions are handled by instances of org.infinispan.commons.dataconversion.Encoder :

public interface Encoder {

   /**
    * Convert data in the read/write format to the storage format.
    *
    * @param content data to be converted, never null.
    * @return Object in the storage format.
    */
   Object toStorage(Object content);

   /**
    * Convert from storage format to the read/write format.
    *
    * @param content data as stored in the cache, never null.
    * @return data in the read/write format
    */
   Object fromStorage(Object content);

   /**
     * Returns the {@link MediaType} produced by this encoder or null if the storage format is not known.
     */
   MediaType getStorageFormat();
}

Data Grid automatically picks the Encoder depending on the cache configuration. The table below shows which internal Encoder is used for several configurations:

It is possible to override programmatically the encoding used for both keys and values, by calling the .withEncoding() method variants from AdvancedCache.

Example, consider the following cache configured as OFF_HEAP:

// Read and write POJO, storage will be byte[] since for
// OFF_HEAP the GlobalMarshallerEncoder is used internally:
cache.put(1, new Pojo())
Pojo value = cache.get(1)

// Get the content in its stored format by overriding
// the internal encoder with a no-op encoder (IdentityEncoder)
Cache<?,?> rawContent = cache.getAdvancedCache().withEncoding(IdentityEncoder.class);
byte[] marshalled = (byte[]) rawContent.get(1);

The override can be useful if any operation in the cache does not require decoding, such as counting number of entries, or calculating the size of byte[] of an OFF_HEAP cache.

A custom encoder can be registered in the EncoderRegistry.

Consider a custom encoder used to compress/decompress with gzip:

public class GzipEncoder implements Encoder {

   @Override
   public Object toStorage(Object content) {
      assert content instanceof String;
      return compress(content.toString());
   }

   @Override
   public Object fromStorage(Object content) {
      assert content instanceof byte[];
      return decompress((byte[]) content);
   }

   private byte[] compress(String str) {
      try (ByteArrayOutputStream baos = new ByteArrayOutputStream();
           GZIPOutputStream gis = new GZIPOutputStream(baos)) {
         gis.write(str.getBytes("UTF-8"));
         gis.close();
         return baos.toByteArray();
      } catch (IOException e) {
         throw new RuntimeException("Unabled to compress", e);
      }
   }

   private String decompress(byte[] compressed) {
      try (GZIPInputStream gis = new GZIPInputStream(new ByteArrayInputStream(compressed));
           BufferedReader bf = new BufferedReader(new InputStreamReader(gis, "UTF-8"))) {
         StringBuilder result = new StringBuilder();
         String line;
         while ((line = bf.readLine()) != null) {
            result.append(line);
         }
         return result.toString();
      } catch (IOException e) {
         throw new RuntimeException("Unable to decompress", e);
      }
   }

   @Override
   public MediaType getStorageFormat() {
      return MediaType.parse("application/gzip");
   }

   @Override
   public boolean isStorageFormatFilterable() {
      return false;
   }

   @Override
   public short id() {
      return 10000;
   }
}

It can be registered by:

GlobalComponentRegistry registry = cacheManager.getGlobalComponentRegistry();
EncoderRegistry encoderRegistry = registry.getComponent(EncoderRegistry.class);
encoderRegistry.registerEncoder(new GzipEncoder());

And then be used to write and read data from a cache:

AdvancedCache<String, String> cache = ...

// Decorate cache with the newly registered encoder, without encoding keys (IdentityEncoder)
// but compressing values
AdvancedCache<String, String> compressingCache = (AdvancedCache<String, String>) cache.withEncoding(IdentityEncoder.class, GzipEncoder.class);

// All values will be stored compressed...
compressingCache.put("297931749", "0412c789a37f5086f743255cfa693dd5");

// ... but API calls deals with String
String stringValue = compressingCache.get("297931749");

// Bypassing the value encoder to obtain the value as it is stored
Object value = compressingCache.withEncoding(IdentityEncoder.class).get("297931749");

// value is a byte[] which is the compressed value

You can deploy and run custom code on Data Grid, such as tasks, listeners, and merge policies. Custom code on Data Grid can directly access data but must also interoperate with clients that access the same data through different endpoints. For example, you can create tasks that handle custom objects while Hot Rod clients read and write data in binary format.

In this case, you can configure application/x-protostream as the cache encoding to store data in binary format then configure your custom code to perform cache operations using a different MediaType.

For example:

DefaultCacheManager cacheManager = new DefaultCacheManager();

// The cache will store POJO for keys and values
ConfigurationBuilder cfg = new ConfigurationBuilder();
cfg.encoding().key().mediaType("application/x-java-object");
cfg.encoding().value().mediaType("application/x-java-object");

cacheManager.defineConfiguration("mycache", cfg.build());

Cache<Integer, Person> cache = cacheManager.getCache("mycache");

cache.put(1, new Person("John","Doe"));

// Wraps cache using 'application/x-java-object' for keys but JSON for values
Cache<Integer, byte[]> jsonValuesCache = (Cache<Integer, byte[]>) cache.getAdvancedCache().withMediaType("application/x-java-object", "application/json");

byte[] json = jsonValuesCache.get(1);

Will return the value in JSON format:

{
   "_type":"org.infinispan.sample.Person",
   "name":"John",
   "surname":"Doe"
}

Data Grid Server includes transcoders by default. However, when running Data Grid as a library, you must add the following to your project:

org.infinispan:infinispan-server-core

Usually there will be none or only one data conversion involved in a cache operation:

But it’s possible to have both encoders and transcoders being used simultaneously for advanced use cases.

Consider an example, a cache that stores marshalled objects (with jboss marshaller) content but for security reasons a transparent encryption layer should be added in order to avoid storing "plain" data to an external store. Clients should be able to read and write data in multiple formats.

This can be achieved by configuring the cache with the the MediaType that describes the storage regardless of the encoding layer:

ConfigurationBuilder cfg = new ConfigurationBuilder();
cfg.encoding().key().mediaType("application/x-jboss-marshalling");
cfg.encoding().key().mediaType("application/x-jboss-marshalling");

The transparent encryption can be added by decorating the cache with a special Encoder that encrypts/decrypts with storing/retrieving, for example:

class Scrambler implements Encoder {

   public Object toStorage(Object content) {
   // Encrypt data
   }

   public Object fromStorage(Object content) {
   // Decrypt data
   }

   @Override
   public boolean isStorageFormatFilterable() {

   }

   public MediaType getStorageFormat() {
   return new MediaType("application", "scrambled");
   }

   @Override
   public short id() {
   //return id
   }
}

To make sure all data written to the cache will be stored encrypted, it’s necessary to decorate the cache with the Encoder above and perform all cache operations in this decorated cache:

Cache<?,?> secureStorageCache = cache.getAdvancedCache().withEncoding(Scrambler.class).put(k,v);

The capability of reading data in multiple formats can be added by decorating the cache with the desired MediaType:

// Obtain a stream of values in XML format from the secure cache
secureStorageCache.getAdvancedCache().withMediaType("application/xml","application/xml").values().stream();

Internally, Data Grid will first apply the encoder fromStorage operation to obtain the entries, that will be in "application/x-jboss-marshalling" format and then apply a successive conversion to "application/xml" by using the adequate Transcoder.

A ProtoStream SerializationContext contains Protobuf type definitions for custom Java objects, loaded from Protobuf schemas, and the accompanying marshallers for those objects.

Data Grid provides a protostream-processor artifact that processes Java annotations in your classes at compile time. The processor generates Protobuf schemas, marshallers, and a concrete implementation of the SerializationContextInitializer interface that you can use to initialize a ProtoStream SerializationContext.

See Registering Serialization Context Initializers.

In some cases you might need to manually define Protobuf schemas and implement ProtoStream marshallers. For example, if you cannot modify Java object classes to add annotations.

See Registering Serialization Context Initializers.

Declare SerializationContextInitializer implementations in your Data Grid configuration to register them.

GlobalConfigurationBuilder builder = new GlobalConfigurationBuilder();
builder.serialization()
       .addContextInitializers(new LibraryInitializerImpl(), new SCIImpl());

<serialization>
    <context-initializer class="org.infinispan.example.LibraryInitializerImpl"/>
    <context-initializer class="org.infinispan.example.another.SCIImpl"/>
</serialization>

JBoss Marshalling is a serialization-based marshalling library and was the default marshaller in previous Data Grid versions.

You can use Java serialization with Data Grid to marshall your objects, but only if your Java objects implement the Java Serializable interface.

Data Grid provides a marshalling implementation that uses Kryo libraries.

Data Grid provides a marshalling implementation that uses Protostuff libraries.

Data Grid provides a Marshaller interface that you can implement for custom marshallers.

Data Grid does not allow deserialization of arbritrary Java classes for security reasons, which applies to JSON, XML, and marshalled byte[] content.

You must add Java classes to a deserialization whitelist, either using system properties or specifying them in the Data Grid configuration.

// Specify a comma-separated list of fully qualified class names
-Dinfinispan.deserialization.whitelist.classes=java.time.Instant,com.myclass.Entity

// Specify a regular expression to match classes
-Dinfinispan.deserialization.whitelist.regexps=.*

<cache-container>
   <serialization version="1.0" marshaller="org.infinispan.marshall.TestObjectStreamMarshaller">
      <white-list>
         <class>org.infinispan.test.data.Person</class>
         <regex>org.infinispan.test.data.*</regex>
       </white-list>
   </serialization>
</cache-container>

To list all the counters defined, the method CounterManager.getCounterNames() returns a collection of all counter names created cluster-wide.

There is a difference between remove a counter via the Strong/WeakCounter interfaces and the CounterManager. The CounterManager.remove(String) removes the counter value from the cluster and removes all the listeners registered in the counter in the local counter instance. In addition, the counter instance is no longer reusable and it may return an invalid results.

On the other side, the Strong/WeakCounter removal only removes the counter value. The instance can still be reused and the listeners still works.

The strong counter provides uses a single key stored in Data Grid cache to provide the consistency needed. All the updates are performed under the key lock to updates its values. On other hand, the reads don’t acquire any locks and reads the current value. Also, with this scheme, it allows to bound the counter value and provide atomic operations like compare-and-set/swap.

A StrongCounter can be retrieved from the CounterManager by using the getStrongCounter() method. As an example:

CounterManager counterManager = ...
StrongCounter aCounter = counterManager.getStrongCounter("my-counter");

The StrongCounter interface adds the following method:

default CompletableFuture<Long> incrementAndGet() {
   return addAndGet(1L);
}

default CompletableFuture<Long> decrementAndGet() {
   return addAndGet(-1L);
}

CompletableFuture<Long> addAndGet(long delta);

CompletableFuture<Boolean> compareAndSet(long expect, long update);

CompletableFuture<Long> compareAndSwap(long expect, long update);

public boolean isUpperBoundReached();
public boolean isLowerBoundReached();

StrongCounter counter = counterManager.getStrongCounter("unbounded_counter");

// incrementing the counter
System.out.println("new value is " + counter.incrementAndGet().get());

// decrement the counter's value by 100 using the functional API
counter.addAndGet(-100).thenApply(v -> {
   System.out.println("new value is " + v);
   return null;
}).get();

// alternative, you can do some work while the counter is updated
CompletableFuture<Long> f = counter.addAndGet(10);
// ... do some work ...
System.out.println("new value is " + f.get());

// and then, check the current value
System.out.println("current value is " + counter.getValue().get());

// finally, reset to initial value
counter.reset().get();
System.out.println("current value is " + counter.getValue().get());

// or set to a new value if zero
System.out.println("compare and set succeeded? " + counter.compareAndSet(0, 1));

StrongCounter counter = counterManager.getStrongCounter("bounded_counter");

// incrementing the counter
try {
    System.out.println("new value is " + counter.addAndGet(100).get());
} catch (ExecutionException e) {
    Throwable cause = e.getCause();
    if (cause instanceof CounterOutOfBoundsException) {
       if (((CounterOutOfBoundsException) cause).isUpperBoundReached()) {
          System.out.println("ops, upper bound reached.");
       } else if (((CounterOutOfBoundsException) cause).isLowerBoundReached()) {
          System.out.println("ops, lower bound reached.");
       }
    }
}

// now using the functional API
counter.addAndGet(-100).handle((v, throwable) -> {
   if (throwable != null) {
      Throwable cause = throwable.getCause();
      if (cause instanceof CounterOutOfBoundsException) {
         if (((CounterOutOfBoundsException) cause).isUpperBoundReached()) {
            System.out.println("ops, upper bound reached.");
         } else if (((CounterOutOfBoundsException) cause).isLowerBoundReached()) {
            System.out.println("ops, lower bound reached.");
         }
      }
      return null;
   }
   System.out.println("new value is " + v);
   return null;
}).get();

StrongCounter counter = counterManager.getStrongCounter("my-counter");
long oldValue, newValue;
do {
   oldValue = counter.getValue().get();
   newValue = someLogic(oldValue);
} while (!counter.compareAndSet(oldValue, newValue).get());

StrongCounter counter = counterManager.getStrongCounter("my-counter");
long oldValue = counter.getValue().get();
long currentValue, newValue;
do {
   currentValue = oldValue;
   newValue = someLogic(oldValue);
} while ((oldValue = counter.compareAndSwap(oldValue, newValue).get()) != currentValue);

The WeakCounter stores the counter’s value in multiple keys in Data Grid cache. The number of keys created is configured by the concurrency-level attribute. Each key stores a partial state of the counter’s value and it can be updated concurrently. It main advantage over the StrongCounter is the lower contention in the cache. On other hand, the read of its value is more expensive and bounds are not allowed.

A WeakCounter can be retrieved from the CounterManager by using the getWeakCounter() method. As an example:

CounterManager counterManager = ...
StrongCounter aCounter = counterManager.getWeakCounter("my-counter);

default CompletableFuture<Void> increment() {
   return add(1L);
}

default CompletableFuture<Void> decrement() {
   return add(-1L);
}

CompletableFuture<Void> add(long delta);

WeakCounter counter = counterManager.getWeakCounter("my_counter");

// increment the counter and check its result
counter.increment().get();
System.out.println("current value is " + counter.getValue());

CompletableFuture<Void> f = counter.add(-100);
//do some work
f.get(); //wait until finished
System.out.println("current value is " + counter.getValue().get());

//using the functional API
counter.reset().whenComplete((aVoid, throwable) -> System.out.println("Reset done " + (throwable == null ? "successfully" : "unsuccessfully"))).get();
System.out.println("current value is " + counter.getValue().get());

In clustered caches, each key has a node responsible to lock the key. This node is called primary owner.

The transactional cache supports optimistic and pessimistic locking mode. Refer to Transaction Locking for more information.

Isolation level affects what transactions can read when running concurrently with other transaction. Refer to Isolation Levels for more information.

The LockManager is a component that is responsible for locking an entry for writing. The LockManager makes use of a LockContainer to locate/hold/create locks. LockContainers come in two broad flavours, with support for lock striping and with support for one lock per entry.

Lock striping entails the use of a fixed-size, shared collection of locks for the entire cache, with locks being allocated to entries based on the entry’s key’s hash code. Similar to the way the JDK’s ConcurrentHashMap allocates locks, this allows for a highly scalable, fixed-overhead locking mechanism in exchange for potentially unrelated entries being blocked by the same lock.

The alternative is to disable lock striping - which would mean a new lock is created per entry. This approach may give you greater concurrent throughput, but it will be at the cost of additional memory usage, garbage collection churn, etc.

The size of the shared lock collection used by lock striping can be tuned using the concurrencyLevel attribute of the <locking /> configuration element.

Configuration example:

<locking striping="false|true"/>

Or

new ConfigurationBuilder().locking().useLockStriping(false|true);

In addition to determining the size of the striped lock container, this concurrency level is also used to tune any JDK ConcurrentHashMap based collections where related, such as internal to DataContainers. Please refer to the JDK ConcurrentHashMap Javadocs for a detailed discussion of concurrency levels, as this parameter is used in exactly the same way in Data Grid.

Configuration example:

<locking concurrency-level="32"/>

Or

new ConfigurationBuilder().locking().concurrencyLevel(32);

The lock timeout specifies the amount of time, in milliseconds, to wait for a contented lock.

Configuration example:

<locking acquire-timeout="10000"/>

Or

new ConfigurationBuilder().locking().lockAcquisitionTimeout(10000);
//alternatively
new ConfigurationBuilder().locking().lockAcquisitionTimeout(10, TimeUnit.SECONDS);

The fact that a single owner is locked (as opposed to all owners being locked) does not break the following consistency guarantee: if key K is hashed to nodes {A, B} and transaction TX1 acquires a lock for K, let’s say on A. If another transaction, TX2, is started on B (or any other node) and TX2 tries to lock K then it will fail with a timeout as the lock is already held by TX1. The reason for this is the that the lock for a key K is always, deterministically, acquired on the same node of the cluster, regardless of where the transaction originates.

From a lock acquisition perspective, pessimistic transactions obtain locks on keys at the time the key is written.

As an example:

transactionManager.begin();
cache.put(k1,v1); //k1 is locked.
cache.remove(k2); //k2 is locked when this returns
transactionManager.commit();

When cache.put(k1,v1) returns, k1 is locked and no other transaction running anywhere in the cluster can write to it. Reading k1 is still possible. The lock on k1 is released when the transaction completes (commits or rollbacks).

With optimistic transactions locks are being acquired at transaction prepare time and are only being held up to the point the transaction commits (or rollbacks). This is different from the 5.0 default locking model where local locks are being acquire on writes and cluster locks are being acquired during prepare time.

As an example:

transactionManager.begin();
cache.put(k1,v1);
cache.remove(k2);
transactionManager.commit(); //at prepare time, K1 and K2 is locked until committed/rolled back.

From a use case perspective, optimistic transactions should be used when there is not a lot of contention between multiple transactions running at the same time. That is because the optimistic transactions rollback if data has changed between the time it was read and the time it was committed (with write skew check enabled).

On the other hand, pessimistic transactions might be a better fit when there is high contention on the keys and transaction rollbacks are less desirable. Pessimistic transactions are more costly by their nature: each write operation potentially involves a RPC for lock acquisition.

To avoid write skews with pessimistic transactions, lock keys at read-time with Flag.FORCE_WRITE_LOCK.

Compare the following code snippets for an example of Flag.FORCE_WRITE_LOCK:

// begin the transaction
if (!cache.getAdvancedCache().lock(key)) {
   // abort the transaction because the key was not locked
} else {
   cache.get(key);
   cache.put(key, value);
   // commit the transaction
}

// begin the transaction
try {
   // throws an exception if the key is not locked.
   cache.getAdvancedCache().withFlags(Flag.FORCE_WRITE_LOCK).get(key);
   cache.put(key, value);
} catch (CacheException e) {
   // mark the transaction rollback-only
}
// commit or rollback the transaction

Once you have configured your cache to use batching, you use it by calling startBatch() and endBatch() on Cache. E.g.,

Cache cache = cacheManager.getCache();
// not using a batch
cache.put("key", "value"); // will replicate immediately

// using a batch
cache.startBatch();
cache.put("k1", "value");
cache.put("k2", "value");
cache.put("k2", "value");
cache.endBatch(true); // This will now replicate the modifications since the batch was started.

// a new batch
cache.startBatch();
cache.put("k1", "value");
cache.put("k2", "value");
cache.put("k3", "value");
cache.endBatch(false); // This will "discard" changes made in the batch

Behind the scenes, the batching functionality starts a JTA transaction, and all the invocations in that scope are associated with it. For this it uses a very simple (e.g. no recovery) internal TransactionManager implementation. With batching, you get:

Consider a distributed transaction in which money is transferred from an account stored in an external database to an account stored in Data Grid. When TransactionManager.commit() is invoked, both resources prepare successfully (1st phase). During the commit (2nd) phase, the database successfully applies the changes whilst Data Grid fails before receiving the commit request from the transaction manager. At this point the system is in an inconsistent state: money is taken from the account in the external database but not visible yet in Data Grid (since locks are only released during 2nd phase of a two-phase commit protocol). Recovery deals with this situation to make sure data in both the database and Data Grid ends up in a consistent state.

Recovery is coordinated by the transaction manager. The transaction manager works with Data Grid to determine the list of in-doubt transactions that require manual intervention and informs the system administrator (via email, log alerts, etc). This process is transaction manager specific, but generally requires some configuration on the transaction manager.  

Knowing the in-doubt transaction ids, the system administrator can now connect to the Data Grid cluster and replay the commit of transactions or force the rollback. Data Grid provides JMX tooling for this - this is explained extensively in the Transaction recovery and reconciliation section.

Recovery is not enabled by default in Data Grid. If disabled, the TransactionManager won’t be able to work with Data Grid to determine the in-doubt transactions. The Transaction configuration section shows how to enable it.

NOTE: recovery-cache attribute is not mandatory and it is configured per-cache.

In order to track in-doubt transactions and be able to reply them, Data Grid caches all transaction state for future use. This state is held only for in-doubt transaction, being removed for successfully completed transactions after when the commit/rollback phase completed.

This in-doubt transaction data is held within a local cache: this allows one to configure swapping this info to disk through cache loader in the case it gets too big. This cache can be specified through the recovery-cache configuration attribute. If not specified Data Grid will configure a local cache for you.

It is possible (though not mandated) to share same recovery cache between all the Data Grid caches that have recovery enabled. If the default recovery cache is overridden, then the specified recovery cache must use a TransactionManagerLookup that returns a different transaction manager than the one used by the cache itself.

Even though this is transaction manager specific, generally a transaction manager would need a reference to a XAResource implementation in order to invoke XAResource.recover() on it. In order to obtain a reference to an Data Grid XAResource following API can be used:

XAResource xar = cache.getAdvancedCache().getXAResource();

It is a common practice to run the recovery in a different process from the one running the transaction.

The transaction manager informs the system administrator on in-doubt transaction in a proprietary way. At this stage it is assumed that the system administrator knows transaction’s XID (a byte array).

A normal recovery flow is:

Figure 10.1. Show in-doubt transactions

The status of each in-doubt transaction is displayed(in this example " PREPARED "). There might be multiple elements in the status field, e.g. "PREPARED" and "COMMITTED" in the case the transaction committed on certain nodes but not on all of them.  

Figure 10.2. Force commit

The recovery design document describes in more detail the insides of transaction recovery implementation.

To enable indexing via XML, you need to add the <indexing> element to your cache configuration, specify the entities that are indexed and optionally pass additional properties.

<infinispan>
   <cache-container default-cache="default">
      <replicated-cache name="default">
         <indexing>
            <indexed-entities>
               <indexed-entity>com.acme.Book</indexed-entity>
            </indexed-entities>
            <property name="property.name">some value</property>
         </indexing>
      </replicated-cache>
   </cache-container>
</infinispan>

import org.infinispan.configuration.cache.*;

ConfigurationBuilder cacheCfg = ...
cacheCfg.indexing().enable()
            .addIndexedEntity(Book.class)
      .addProperty("property name", "propery value")

It is recommended to declare the indexed types, as they will be mandatory in the next Data Grid version.

<infinispan>
   <cache-container default-cache="default">
      <replicated-cache name="default">
         <indexing>
            <indexed-entities>
                <indexed-entity>com.acme.query.test.Car</indexed-entity>
                <indexed-entity>com.acme.query.test.Truck</indexed-entity>
            </indexed-entities>
         </indexing>
      </replicated-cache>
   </cache-container>
</infinispan>

 cacheCfg.indexing()
       .addIndexedEntity(Car.class)
       .addIndexedEntity(Truck.class)

When the cache is storing protobuf, the indexed types should be the Message declared in the protobuf schema. For example, for the schema below:

package book_sample;

message Book {
    optional string title = 1;
    optional string description = 2;
    optional int32 publicationYear = 3; // no native Date type available in Protobuf

    repeated Author authors = 4;
}

message Author {
    optional string name = 1;
    optional string surname = 2;
}

The config should be:

<infinispan>
  <cache-container default-cache="default">
    <replicated-cache name="books">
      <indexing>
        <indexed-entities>
          <indexed-entity>book_sample.Book</indexed-entity>
        </indexed-entities>
      </indexing>
    </replicated-cache>
  </cache-container>
</infinispan>

Data Grid can store indexes in the file system or in memory (local-heap). File system is the recommended and the default configuration, and memory indexes should only be used for small to medium indexes that don’t need to survive restart.

<replicated-cache name="myCache">
   <indexing>
      <indexed-entities>
         <indexed-entity>com.acme.Book</indexed-entity>
      </indexed-entities>
      <!-- Optional: this is the default setting -->
      <property name="default.directory_provider">filesystem</property>
      <!-- Optional: define base folder for indexes -->
      <property name="default.indexBase">${java.io.tmpdir}/baseDir</property>
   </indexing>
</replicated-cache>

<replicated-cache name="myCache">
   <indexing>
      <indexed-entities>
         <indexed-entity>com.acme.Book</indexed-entity>
      </indexed-entities>
      <property name="default.directory_provider">local-heap</property>
   </indexing>
</replicated-cache>

Data Grid uses internally a component called "Index Manager" to control how new data is applied to the index and when the data is visible to searches.

The default Index Manager directory-based writes to the index as soon as the data is written to the cache. The downside is it can slow down considerably cache writes specially under heavy writing scenarios, since it needs to do constant costly operations called "flushes" on the index.

The near-real-time index manager is similar to the default index manager but takes advantage of the Near-Real-Time features of Lucene. It has better write performance because it flushes the index to the underlying store less often. The drawback is that unflushed index changes can be lost in case of a non-clean shutdown. Can be used in conjunction with local-heap or filesystem.

Example with local-heap:

<replicated-cache name="default">
    <indexing>
        <property name="default.indexmanager">near-real-time</property>
        <property name="default.directory_provider">local-heap</property>
    </indexing>
</replicated-cache>

Example with filesystem:

<replicated-cache name="default">
    <indexing>
        <property name="default.indexmanager">near-real-time</property>
    </indexing>
</replicated-cache>

Rebuilding an index reconstructs it from data stored in the cache. You need to rebuild indexes if you change things like definitions of indexed types or Analyzers. Likewise you might need to rebuild indexes if they are deleted for some reason. Beware it might take some time as it needs to reprocess all data in the grid!

Indexer indexer = Search.getIndexer(cache);
CompletionStage<Void> future = index.run();

You can limit the number of returned results by using the Query.maxResults(int maxResults). This can be used in conjunction with Query.startOffset(long startOffset) to achieve pagination of the result set.

// sorted by year and match all books that have "clustering" in their title
// and return the third page of 10 results
Query<Book> query = queryFactory.create("FROM com.acme.Book WHERE title like '%clustering%' ORDER BY year").startOffset(20).maxResults(10)

The QueryResult object has the .hitCount() method to return the total number of results of the query, regardless of any pagination parameter. The hit count is only available for indexed queries for performance reasons.

The Query object has the .iterator() method to obtain the results lazily. It returns an instance of CloseableIterator that must be closed after usage.

Instead of building a new Query object for every execution it is possible to include named parameters in the query which can be substituted with actual values before execution. This allows a query to be defined once and be efficiently executed many times. Parameters can only be used on the right-hand side of an operator and are defined when the query is created by supplying an object produced by the org.infinispan.query.dsl.Expression.param(String paramName) method to the operator instead of the usual constant value. Once the parameters have been defined they can be set by invoking either Query.setParameter(parameterName, value) or Query.setParameters(parameterMap) as shown in the examples below. ⁠

QueryFactory queryFactory = Search.getQueryFactory(cache);
// Defining a query to search for various authors and publication years
Query<Book> query = queryFactory.create("SELECT title FROM com.acme.Book WHERE author = :authorName AND publicationYear = :publicationYear").build();

// Set actual parameter values
query.setParameter("authorName", "Doe");
query.setParameter("publicationYear", 2010);

// Execute the query
List<Book> found = query.list();

Alternatively, you can supply a map of actual parameter values to set multiple parameters at once: ⁠

Map<String, Object> parameterMap = new HashMap<>();
parameterMap.put("authorName", "Doe");
parameterMap.put("publicationYear", 2010);

query.setParameters(parameterMap);

The Ickle query language is small subset of the JPQL query language, with some extensions for full-text.

The parser syntax has some notable rules:

# match all books that have "Data Grid" in their title
# or have an author named "Manik" and their description contains "clustering"

FROM com.acme.Book WHERE title LIKE '%Data Grid%' OR author.name = 'Manik' AND description like '%clustering%'

# match all books that do not have "Data Grid" in their title and are authored by "Manik"
FROM com.acme.Book WHERE title != 'Data Grid' AND author.name = 'Manik'

# match all books that have an author named "Manik" and their title contains
# "Data Grid" or their description contains "clustering"
FROM com.acme.Book WHERE author.name = 'Manik' AND ( title like '%Data Grid%' OR description like '% clustering%')

# match all books that have "Data Grid" in their title or description
# and return only their title and publication year
SELECT title, publicationYear FROM com.acme.Book WHERE title like '%Data Grid%' OR description like '%Data Grid%'

# match all books that have "Data Grid" in their title or description
# and return them sorted by the publication year and title
FROM com.acme.Book WHERE title like '%Data Grid%' ORDER BY publicationYear DESC, title ASC

SELECT author, COUNT(title) FROM com.acme.Book WHERE title LIKE '%engine%' GROUP BY author

FROM sample_bank_account.Transaction WHERE description : 'cofee'~2

FROM sample_bank_account.Transaction WHERE amount : [20 to 50]

FROM sample_bank_account.Transaction WHERE description : 'bus fare'

FROM sample_bank_account.Transaction WHERE description : 'canceling fee'~3

FROM sample_bank_account.Transaction where description : 'te?t'

FROM sample_bank_account.Transaction where description : 'test*'

FROM sample_library.Book  where title : /[mb]oat/

FROM sample_library.Book WHERE title : beer^3 OR wine

We’re going to store Book instances in an Data Grid cache called "books". Book instances will be indexed, so we enable indexing for the cache:

Data Grid configuration:

<infinispan>
    <cache-container>
        <transport cluster="infinispan-cluster"/>
        <distributed-cache name="books">
            <indexing>
                <indexed-entities>
                    <indexed-entity>com.acme.Book</indexed-entity>
                </indexed-entities>
            </indexing>
        </distributed-cache>
    </cache-container>
</infinispan>

Obtaining the cache:

import org.infinispan.Cache;
import org.infinispan.manager.DefaultCacheManager;
import org.infinispan.manager.EmbeddedCacheManager;

EmbeddedCacheManager manager = new DefaultCacheManager("infinispan.xml");
Cache<String, Book> cache = manager.getCache("books");

Each Book will be defined as in the following example; we have to choose which properties are indexed, and for each property we can optionally choose advanced indexing options using the annotations defined in the Hibernate Search project.

import org.hibernate.search.annotations.*;
import java.util.Date;
import java.util.HashSet;
import java.util.Set;

//Values you want to index need to be annotated with @Indexed, then you pick which fields and how they are to be indexed:
@Indexed
public class Book {
   @Field String title;
   @Field String description;
   @Field @DateBridge(resolution=Resolution.YEAR) Date publicationYear;
   @IndexedEmbedded Set<Author> authors = new HashSet<Author>();
}

public class Author {
   @Field String name;
   @Field String surname;
   // hashCode() and equals() omitted
}

Now assuming we stored several Book instances in our Data Grid Cache , we can search them for any matching field as in the following example.

// get the query factory from the cache:
QueryFactory queryFactory = org.infinispan.query.Search.getQueryFactory(cache);

// create an Ickle query that will do a full-text search (operator ':') on fields 'title' and 'authors.name'
Query<Book> fullTextQuery = queryFactory.create("FROM com.acme.Book WHERE title:'infinispan' AND authors.name:'sanne'")

// The ('=') operator is not a full-text operator, thus can be used in both indexed and non-indexed caches
Query<Book> exactMatchQuery = queryFactory.create("FROM com.acme.Book WHERE title = 'Programming Infinispan' AND authors.name = 'Sanne Grinnovero'")

// Full-text and non-full text operators can be part of the same query
Query q = queryFactory.create("FROM com.query.Book b where b.author.name = 'Stephen' and b.description : (+'dark' -'tower')");

// get the results
List<Book> found=query.execute().list();

Apart from list() you have the option for obtaining on iterator(), or use pagination.

Data Grid relies on the rich API of Hibernate Search in order to define fine grained configuration for indexing at entity level. This configuration includes which fields are annotated, which analyzers should be used, how to map nested objects and so on. Detailed documentation is available at the Hibernate Search manual.

@Transformable(transformer = CustomTransformer.class)
public class CustomKey {
   ...
}

public class CustomTransformer implements Transformer {
   @Override
   public Object fromString(String s) {
      ...
      return new CustomKey(...);
   }

   @Override
   public String toString(Object customType) {
      CustomKey ck = (CustomKey) customType;
      return ...
   }
}

<replicated-cache name="test">
    <indexing auto-config="true">
        <key-transformers>
            <key-transformer key="com.mycompany.CustomKey" transformer="com.mycompany.CustomTransformer"/>
        </key-transformers>
    </indexing>
</replicated-cache>

   ConfigurationBuilder builder = ...
   builder.indexing().enable()
         .addKeyTransformer(CustomKey.class, CustomTransformer.class);

import org.apache.lucene.search.Query;
import org.hibernate.search.cfg.Environment;
import org.hibernate.search.cfg.SearchMapping;
import org.hibernate.search.query.dsl.QueryBuilder;
import org.infinispan.Cache;
import org.infinispan.configuration.cache.Configuration;
import org.infinispan.configuration.cache.ConfigurationBuilder;
import org.infinispan.configuration.cache.Index;
import org.infinispan.manager.DefaultCacheManager;
import org.infinispan.query.CacheQuery;
import org.infinispan.query.Search;
import org.infinispan.query.SearchManager;

import java.io.IOException;
import java.lang.annotation.ElementType;
import java.util.Properties;

SearchMapping mapping = new SearchMapping();
mapping.entity(Author.class).indexed()
       .property("name", ElementType.METHOD).field()
       .property("surname", ElementType.METHOD).field();

Properties properties = new Properties();
properties.put(Environment.MODEL_MAPPING, mapping);
properties.put("hibernate.search.[other options]", "[...]");

Configuration infinispanConfiguration = new ConfigurationBuilder()
        .indexing().index(Index.NONE)
        .withProperties(properties)
        .build();

DefaultCacheManager cacheManager = new DefaultCacheManager(infinispanConfiguration);

Cache<Long, Author> cache = cacheManager.getCache();
SearchManager sm = Search.getSearchManager(cache);

Author author = new Author(1, "Manik", "Surtani");
cache.put(author.getId(), author);

QueryBuilder qb = sm.buildQueryBuilderForClass(Author.class).get();
Query q = qb.keyword().onField("name").matching("Manik").createQuery();
CacheQuery cq = sm.getQuery(q, Author.class);
assert cq.getResultSize() == 1;

We are going to revisit the Book Sample from embedded query, but this time using the Java Hot Rod client and the Infinispan server. An object called Book will be stored in a Infinispan cache called "books". Book instances will be indexed, so we enable indexing for the cache:

<infinispan>
  <cache-container default-cache="default">
    <replicated-cache name="books">
      <indexing>
        <indexed-entities>
          <indexed-entity>book_sample.Book</indexed-entity>
        </indexed-entities>
      </indexing>
    </replicated-cache>
  </cache-container>
</infinispan>

Alternatively, if indexing the cache is not indexed, we configure the <encoding> as application/x-protostream to make sure the storage is queryable:

<infinispan>
  <cache-container default-cache="default">
    <replicated-cache name="books">
      <encoding media-type="application/x-protostream"/>
    </replicated-cache>
  </cache-container>
</infinispan>

Each Book will be defined as in the following example: we use @Protofield annotations to identify the protocol buffers message fields and the @ProtoDoc annotation on the fields to configure indexing attributes:

import org.infinispan.protostream.annotations.ProtoDoc;
import org.infinispan.protostream.annotations.ProtoFactory;
import org.infinispan.protostream.annotations.ProtoField;

@ProtoDoc("@Indexed")
public class Book {

   @ProtoDoc("@Field(index=Index.YES, analyze = Analyze.YES, store = Store.NO)")
   @ProtoField(number = 1)
   final String title;

   @ProtoDoc("@Field(index=Index.YES, analyze = Analyze.YES, store = Store.NO)")
   @ProtoField(number = 2)
   final String description;

   @ProtoDoc("@Field(index=Index.YES, analyze = Analyze.YES, store = Store.NO)")
   @ProtoField(number = 3, defaultValue = "0")
   final int publicationYear;


   @ProtoFactory
   Book(String title, String description, int publicationYear) {
      this.title = title;
      this.description = description;
      this.publicationYear = publicationYear;
   }
   // public Getter methods omitted for brevity
}

The annotations above will generate during compilation the artifacts necessary to read, write and query Book instances. To enable this generation, use the @AutoProtoSchemaBuilder annotation in a newly created class with empty constructor or interface:

import org.infinispan.protostream.SerializationContextInitializer;
import org.infinispan.protostream.annotations.AutoProtoSchemaBuilder;

@AutoProtoSchemaBuilder(
      includeClasses = {
            Book.class
      },
      schemaFileName = "book.proto",
      schemaFilePath = "proto/",
      schemaPackageName = "book_sample")
public interface RemoteQueryInitializer extends SerializationContextInitializer {
}

After compilation, a file book.proto file will be created in the configured schemaFilePath, along with an implementation RemoteQueryInitializerImpl.java of the annotated interface. This concrete class can be used directly in the Hot Rod client code to initialize the serialization context.

Putting all together:

package org.infinispan;

import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.List;

import org.infinispan.client.hotrod.RemoteCache;
import org.infinispan.client.hotrod.RemoteCacheManager;
import org.infinispan.client.hotrod.Search;
import org.infinispan.client.hotrod.configuration.ConfigurationBuilder;
import org.infinispan.query.dsl.Query;
import org.infinispan.query.dsl.QueryFactory;
import org.infinispan.query.remote.client.ProtobufMetadataManagerConstants;

public class RemoteQuery {

   public static void main(String[] args) throws Exception {
      ConfigurationBuilder clientBuilder = new ConfigurationBuilder();
      // RemoteQueryInitializerImpl is generated
      clientBuilder.addServer().host("127.0.0.1").port(11222)
            .security().authentication().username("user").password("user")
            .addContextInitializers(new RemoteQueryInitializerImpl());

      RemoteCacheManager remoteCacheManager = new RemoteCacheManager(clientBuilder.build());

      // Grab the generated protobuf schema and registers in the server.
      Path proto = Paths.get(RemoteQuery.class.getClassLoader()
            .getResource("proto/book.proto").toURI());
      String protoBufCacheName = ProtobufMetadataManagerConstants.PROTOBUF_METADATA_CACHE_NAME;
      remoteCacheManager.getCache(protoBufCacheName).put("book.proto", Files.readString(proto));

      // Obtain the 'books' remote cache
      RemoteCache<Object, Object> remoteCache = remoteCacheManager.getCache("books");

      // Add some Books
      Book book1 = new Book("Infinispan in Action", "Learn Infinispan with using it", 2015);
      Book book2 = new Book("Cloud-Native Applications with Java and Quarkus", "Build robust and reliable cloud applications", 2019);

      remoteCache.put(1, book1);
      remoteCache.put(2, book2);

      // Execute a full-text query
      QueryFactory queryFactory = Search.getQueryFactory(remoteCache);
      Query<Book> query = queryFactory.create("FROM book_sample.Book WHERE title:'java'");

      List<Book> list = query.execute().list(); // Voila! We have our book back from the cache!
   }
}

As seen in Remote Query example, one step necessary to query protobuf entities is to provide the client and server with the relevant metadata about entities (.proto file).

The descriptors are stored in a dedicated cache on the server named ___protobuf_metadata. Both keys and values in this cache are plain strings. Registering a new schema is therefore as simple as performing a put() operation on this cache using the schema’s name as key and the schema file itself as the value.

Alternatively you can use the CLI (via the cache-container=*:register-proto-schemas() operation), the Management Console, the REST endpoint /rest/v2/schemas or the ProtobufMetadataManager MBean via JMX. Be aware that, when security is enabled, access to the schema cache via the remote protocols requires that the user belongs to the '___schema_manager' role.

Analysis is a process that converts input data into one or more terms that you can index and query. While in Embedded Query mapping is done through Hibernate Search annotations, that supports a rich set of Lucene based analyzers, in client-server mode the analyzer definitions are declared in a platform neutral way

/* @Indexed */
message TestEntity {

    /* @Field(store = Store.YES, analyze = Analyze.YES, analyzer = @Analyzer(definition = "keyword")) */
    optional string id = 1;

    /* @Field(store = Store.YES, analyze = Analyze.YES, analyzer = @Analyzer(definition = "simple")) */
    optional string name = 2;
}

@ProtoDoc("@Field(store = Store.YES, analyze = Analyze.YES, analyzer = @Analyzer(definition = \"keyword\"))")
@ProtoField(number = 1)
final String id;

@ProtoDoc("@Field(store = Store.YES, analyze = Analyze.YES, analyzer = @Analyzer(definition = \"simple\"))")
@ProtoField(number = 2)
final String description;

A continuous query uses a listener that is notified when:

When a client registers a continuous query listener it immediately begins to receive the results currently matching the query, received as Join events as described above. In addition, it will receive subsequent notifications when other entries begin matching the query, as Join events, or stop matching the query, as Leave events, as a consequence of any cache operations that would normally generate creation, modification, removal, or expiration events. Updated cache entries will generate Update events if the entry matches the query filter before and after the operation. To summarize, the logic used to determine if the listener receives a Join, Update or Leave event is:

To create a continuous query, do the following:

continuousQuery.addContinuousQueryListener(query, listener);

The following example demonstrates a simple continuous query use case in embedded mode: ⁠

import org.infinispan.query.api.continuous.ContinuousQuery;
import org.infinispan.query.api.continuous.ContinuousQueryListener;
import org.infinispan.query.Search;
import org.infinispan.query.dsl.QueryFactory;
import org.infinispan.query.dsl.Query;

import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;

[...]

// We have a cache of Persons
Cache<Integer, Person> cache = ...

// We begin by creating a ContinuousQuery instance on the cache
ContinuousQuery<Integer, Person> continuousQuery = Search.getContinuousQuery(cache);

// Define our query. In this case we will be looking for any Person instances under 21 years of age.
QueryFactory queryFactory = Search.getQueryFactory(cache);
Query query = queryFactory.create("FROM Person p WHERE p.age < 21");

final Map<Integer, Person> matches = new ConcurrentHashMap<Integer, Person>();

// Define the ContinuousQueryListener
ContinuousQueryListener<Integer, Person> listener = new ContinuousQueryListener<Integer, Person>() {
    @Override
    public void resultJoining(Integer key, Person value) {
        matches.put(key, value);
    }

    @Override
    public void resultUpdated(Integer key, Person value) {
        // we do not process this event
    }

    @Override
    public void resultLeaving(Integer key) {
        matches.remove(key);
    }
};

// Add the listener and the query
continuousQuery.addContinuousQueryListener(query, listener);

[...]

// Remove the listener to stop receiving notifications
continuousQuery.removeContinuousQueryListener(listener);

As Person instances having an age less than 21 are added to the cache they will be received by the listener and will be placed into the matches map, and when these entries are removed from the cache or their age is modified to be greater or equal than 21 they will be removed from matches.

To stop the query from further execution just remove the listener:

continuousQuery.removeContinuousQueryListener(listener);

Continuous queries are designed to provide a constant stream of updates to the application, potentially resulting in a very large number of events being generated for particularly broad queries. A new temporary memory allocation is made for each event. This behavior may result in memory pressure, potentially leading to OutOfMemoryErrors (especially in remote mode) if queries are not carefully designed. To prevent such issues it is strongly recommended to ensure that each query captures the minimal information needed both in terms of number of matched entries and size of each match (projections can be used to capture the interesting properties), and that each ContinuousQueryListener is designed to quickly process all received events without blocking and to avoid performing actions that will lead to the generation of new matching events from the cache it listens to.

By default, the Index Managers work in sync mode, meaning when data is written to Data Grid, it will perform the indexing operations synchronously. This synchronicity guarantees indexes are always consistent with the data (and thus visible in searches), but can slowdown write operations since it will also perform a commit to the index. Committing is an extremely expensive operation in Lucene, and for that reason, multiple writes from different nodes can be automatically batched into a single commit to reduce the impact.

So, when doing data loads to Data Grid with index enabled, try to use multiple threads to take advantage of this batching.

If using multiple threads does not result in the required performance, an alternative is to load data with indexing temporarily disabled and run a re-indexing operation afterwards. This can be done writing data with the SKIP_INDEXING flag:

cache.getAdvancedCache().withFlags(Flag.SKIP_INDEXING).put("key","value");

If it’s acceptable a small delay between data writes and when that data is visible in queries, an index manager can be configured to work in async mode. The async mode offers much better writing performance, since in this mode commits happen at a configurable interval.

Configuration:

<distributed-cache name="default">
    <indexing>
        <!-- Index data in async mode -->
        <property name="default.worker.execution">async</property>
        <!-- Optional: configure the commit interval, default is 1000ms -->
        <property name="default.index_flush_interval">500</property>
    </indexing>
</distributed-cache>

Lucene internally works with snapshots of the index: once an IndexReader is opened, it will only see the index changes up to the point it was opened; further index changes will not be visible until the IndexReader is refreshed. The Index Managers used in Data Grid by default will check the freshness of the index readers before every query and refresh them if necessary.

It is possible to tune this strategy to relax this freshness checking to a pre-configured interval by using the reader.strategy configuration set as async:

<distributed-cache name="default">
    <indexing>
        <property name="default.reader.strategy">async</property>
        <!-- refresh reader every 1s, default is 5s -->
        <property name="default.reader.async_refresh_period_ms">1000</property>
    </indexing>
</distributed-cache>

It is possible to apply tuning options in Lucene directly. For more details, see the Hibernate Search manual.

It is possible to limit on which nodes the command will be ran. For example you may want to only run a computation on machines in the same rack. Or you may want to perform an operation once in the local site and again on a different site. A cluster executor can limit what nodes it sends requests to at the scope of same or different machine, rack or site level.

EmbeddedCacheManager manager = ...;
manager.executor().filterTargets(ClusterExecutionPolicy.SAME_RACK).submit(...)

To use this topology base filtering you must enable topology aware consistent hashing through Server Hinting.

You can also filter using a predicate based on the Address of the node. This can also be optionally combined with topology based filtering in the previous code snippet.

We also allow the target node to be chosen by any means using a Predicate that will filter out which nodes can be considered for execution. Note this can also be combined with Topology filtering at the same time to allow even more fine control of where you code is executed within the cluster.

EmbeddedCacheManager manager = ...;
// Just filter
manager.executor().filterTargets(a -> a.equals(..)).submit(...)
// Filter only those in the desired topology
manager.executor().filterTargets(ClusterExecutionPolicy.SAME_SITE, a -> a.equals(..)).submit(...)

Cluster Executor allows for a timeout to be set per invocation. This defaults to the distributed sync timeout as configured on the Transport Configuration. This timeout works in both a clustered and non clustered cache manager. The executor may or may not interrupt the threads executing a task when the timeout expires. However when the timeout occurs any Consumer or Future will be completed passing back a TimeoutException. This value can be overridden by ivoking the timeout method and supplying the desired duration.

Cluster Executor can also run in single node submission mode instead of submitting the command to all nodes it will instead pick one of the nodes that would have normally received the command and instead submit it it to only one. Each submission will possibly use a different node to execute the task on. This can be very useful to use the ClusterExecutor as a java.util.concurrent.Executor which you may have noticed that ClusterExecutor implements.

EmbeddedCacheManager manager = ...;
manager.executor().singleNodeSubmission().submit(...)

This example shows how you can use the ClusterExecutor to estimate the value of PI.

Pi approximation can greatly benefit from parallel distributed execution via Cluster Executor. Recall that area of the square is Sa = 4r2 and area of the circle is Ca=pi*r2. Substituting r2 from the second equation into the first one it turns out that pi = 4 * Ca/Sa. Now, image that we can shoot very large number of darts into a square; if we take ratio of darts that land inside a circle over a total number of darts shot we will approximate Ca/Sa value. Since we know that pi = 4 * Ca/Sa we can easily derive approximate value of pi. The more darts we shoot the better approximation we get. In the example below we shoot 1 billion darts but instead of "shooting" them serially we parallelize work of dart shooting across the entire Data Grid cluster. Note this will work in a cluster of 1 was well, but will be slower.

public class PiAppx {

   public static void main (String [] arg){
      EmbeddedCacheManager cacheManager = ..
      boolean isCluster = ..

      int numPoints = 1_000_000_000;
      int numServers = isCluster ? cacheManager.getMembers().size() : 1;
      int numberPerWorker = numPoints / numServers;

      ClusterExecutor clusterExecutor = cacheManager.executor();
      long start = System.currentTimeMillis();
      // We receive results concurrently - need to handle that
      AtomicLong countCircle = new AtomicLong();
      CompletableFuture<Void> fut = clusterExecutor.submitConsumer(m -> {
         int insideCircleCount = 0;
         for (int i = 0; i < numberPerWorker; i++) {
            double x = Math.random();
            double y = Math.random();
            if (insideCircle(x, y))
               insideCircleCount++;
         }
         return insideCircleCount;
      }, (address, count, throwable) -> {
         if (throwable != null) {
            throwable.printStackTrace();
            System.out.println("Address: " + address + " encountered an error: " + throwable);
         } else {
            countCircle.getAndAdd(count);
         }
      });
      fut.whenComplete((v, t) -> {
         // This is invoked after all nodes have responded with a value or exception
         if (t != null) {
            t.printStackTrace();
            System.out.println("Exception encountered while waiting:" + t);
         } else {
            double appxPi = 4.0 * countCircle.get() / numPoints;

            System.out.println("Distributed PI appx is " + appxPi +
                  " using " + numServers + " node(s), completed in " + (System.currentTimeMillis() - start) + " ms");
         }
      });

      // May have to sleep here to keep alive if no user threads left
   }

   private static boolean insideCircle(double x, double y) {
      return (Math.pow(x - 0.5, 2) + Math.pow(y - 0.5, 2))
            <= Math.pow(0.5, 2);
   }
}

Internally the data is segmented and each node only performs the operations upon the data it owns as a primary owner. This allows for data to be processed evenly, assuming segments are granular enough to provide for equal amounts of data on each node.

When you are utilizing a distributed cache, the data can be reshuffled between nodes when a new node joins or leaves. Distributed Streams handle this reshuffling of data automatically so you don’t have to worry about monitoring when nodes leave or join the cluster. Reshuffled entries may be processed a second time, and we keep track of the processed entries at the key level or at the segment level (depending on the terminal operation) to limit the amount of duplicate processing.

It is possible but highly discouraged to disable rehash awareness on the stream. This should only be considered if your request can handle only seeing a subset of data if a rehash occurs. This can be done by invoking CacheStream.disableRehashAware() The performance gain for most operations when a rehash doesn’t occur is completely negligible. The only exceptions are for iterator and forEach, which will use less memory, since they do not have to keep track of processed keys.

Since the operations are sent across to other nodes they must be serializable by Data Grid marshalling. This allows the operations to be sent to the other nodes.

The simplest way is to use a CacheStream instance and use a lambda just as you would normally. Data Grid overrides all of the various Stream intermediate and terminal methods to take Serializable versions of the arguments (ie. SerializableFunction, SerializablePredicate…​) You can find these methods at CacheStream. This relies on the spec to pick the most specific method as defined here.

In our previous example we used a Collector to collect all the results into a Map. Unfortunately the Collectors class doesn’t produce Serializable instances. Thus if you need to use these, there are two ways to do so:

One option would be to use the CacheCollectors class which allows for a Supplier<Collector> to be provided. This instance could then use the Collectors to supply a Collector which is not serialized.

Map<Object, String> jbossValues = cache.entrySet().stream()
              .filter(e -> e.getValue().contains("Jboss"))
              .collect(CacheCollectors.serializableCollector(() -> Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue)));

Alternatively, you can avoid the use of CacheCollectors and instead use the overloaded collect methods that take Supplier<Collector>. These overloaded collect methods are only available via CacheStream interface.

Map<Object, String> jbossValues = cache.entrySet().stream()
              .filter(e -> e.getValue().contains("Jboss"))
              .collect(() -> Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue));

If however you are not able to use the Cache and CacheStream interfaces you cannot utilize Serializable arguments and you must instead cast the lambdas to be Serializable manually by casting the lambda to multiple interfaces. It is not a pretty sight but it gets the job done.

Map<Object, String> jbossValues = map.entrySet().stream()
              .filter((Serializable & Predicate<Map.Entry<Object, String>>) e -> e.getValue().contains("Jboss"))
              .collect(CacheCollectors.serializableCollector(() -> Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue)));

The recommended and most performant way is to use an AdvancedExternalizer as this provides the smallest payload. Unfortunately this means you cannot use lamdbas as advanced externalizers require defining the class before hand.

You can use an advanced externalizer as shown below:

   Map<Object, String> jbossValues = cache.entrySet().stream()
              .filter(new ContainsFilter("Jboss"))
              .collect(() -> Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue));

   class ContainsFilter implements Predicate<Map.Entry<Object, String>> {
      private final String target;

      ContainsFilter(String target) {
         this.target = target;
      }

      @Override
      public boolean test(Map.Entry<Object, String> e) {
         return e.getValue().contains(target);
      }
   }

   class JbossFilterExternalizer implements AdvancedExternalizer<ContainsFilter> {

      @Override
      public Set<Class<? extends ContainsFilter>> getTypeClasses() {
         return Util.asSet(ContainsFilter.class);
      }

      @Override
      public Integer getId() {
         return CUSTOM_ID;
      }

      @Override
      public void writeObject(ObjectOutput output, ContainsFilter object) throws IOException {
         output.writeUTF(object.target);
      }

      @Override
      public ContainsFilter readObject(ObjectInput input) throws IOException, ClassNotFoundException {
         return new ContainsFilter(input.readUTF());
      }
   }

You could also use an advanced externalizer for the collector supplier to reduce the payload size even further.

Map<Object, String> map = (Map<Object, String>) cache.entrySet().stream()
              .filter(new ContainsFilter("Jboss"))
              .collect(Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue));

 class ToMapCollectorSupplier<K, U> implements Supplier<Collector<Map.Entry<K, U>, ?, Map<K, U>>> {
      static final ToMapCollectorSupplier INSTANCE = new ToMapCollectorSupplier();

      private ToMapCollectorSupplier() { }

      @Override
      public Collector<Map.Entry<K, U>, ?, Map<K, U>> get() {
         return Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue);
      }
   }

   class ToMapCollectorSupplierExternalizer implements AdvancedExternalizer<ToMapCollectorSupplier> {

      @Override
      public Set<Class<? extends ToMapCollectorSupplier>> getTypeClasses() {
         return Util.asSet(ToMapCollectorSupplier.class);
      }

      @Override
      public Integer getId() {
         return CUSTOM_ID;
      }

      @Override
      public void writeObject(ObjectOutput output, ToMapCollectorSupplier object) throws IOException {
      }

      @Override
      public ToMapCollectorSupplier readObject(ObjectInput input) throws IOException, ClassNotFoundException {
         return ToMapCollectorSupplier.INSTANCE;
      }
   }

import java.net.URI;
import javax.cache.*;
import javax.cache.configuration.*;

// Load configuration from an absolute filesystem path
URI uri = URI.create("file:///path/to/infinispan.xml");
// Load configuration from a classpath resource
// URI uri = this.getClass().getClassLoader().getResource("infinispan.xml").toURI();

// Create a cache manager using the above configuration
CacheManager cacheManager = Caching.getCachingProvider().getCacheManager(uri, this.getClass().getClassLoader(), null);

Alternatively, JCache can be configured to store data by reference (just like Data Grid or JDK Collections work). To do that, simply call:

Cache<String, String> cache = cacheManager.createCache("namedCache",
      new MutableConfiguration<String, String>().setStoreByValue(false));

Hot Rod configuration files include infinispan.client.hotrod.cache.* properties that you can use to customize remote caches.

import javax.cache.*;
import javax.cache.configuration.*;

// Load configuration from an absolute filesystem path
URI uri = URI.create("file:///path/to/hotrod-client.properties");
// Load configuration from a classpath resource
// URI uri = this.getClass().getClassLoader().getResource("hotrod-client.properties").toURI();

// Retrieve the system wide cache manager via org.infinispan.jcache.remote.JCachingProvider
CacheManager cacheManager = Caching.getCachingProvider("org.infinispan.jcache.remote.JCachingProvider")
      .getCacheManager(uri, this.getClass().getClassLoader(), null);

// create or obtain your EmbeddedCacheManager
EmbeddedCacheManager cm = ... ;

// create or obtain a MultimapCacheManager passing the EmbeddedCacheManager
MultimapCacheManager multimapCacheManager = EmbeddedMultimapCacheManagerFactory.from(cm);

// define the configuration for the multimap cache
multimapCacheManager.defineConfiguration(multimapCacheName, c.build());

// get the multimap cache
multimapCache = multimapCacheManager.get(multimapCacheName);

Duplicates are not supported yet. This means that the multimap won’t contain any duplicate key-value pair. Whenever put method is called, if the key-value pair already exist, this key-value par won’t be added. Methods used to check if a key-value pair is already present in the Multimap are the equals and hashcode.

For now, the eviction works per key, and not per key-value pair. This means that whenever a key is evicted, all the values associated with the key will be evicted too.

Implicit transactions are supported through the auto-commit and all the methods are non blocking. Explicit transactions work without blocking in most of the cases. Methods that will block are size, containsEntry and remove(Predicate<? super V> p)

The following table lists and describes configuration properties for the HotRodManager class: