Chapter 6. Clustering

   <local-cache name="mySimpleCache" simple-cache="true">
        <!-- expiration, eviction, security... -->
   </local-cache>

CacheManager cm = getCacheManager();
ConfigurationBuilder builder = new ConfigurationBuilder().simpleCache(true);
cm.defineConfiguration("mySimpleCache", builder.build());
Cache cache = cm.getCache("mySimpleCache");

Simple cache checks against features it does not support, if you configure it to use e.g. transactions, configuration validation will throw an exception.

Capacity factors allocate segment-to-node mappings based on resources available to nodes.

To configure capacity factors, you specify any non-negative number and the Red Hat Data Grid hashing algorithm assigns each node a load weighted by its capacity factor (both as a primary owner and as a backup owner).

For example, nodeA has 2x the memory available than nodeB in the same Red Hat Data Grid cluster. In this case, setting capacityFactor to a value of 2 configures Red Hat Data Grid to allocate 2x the number of segments to nodeA.

Setting a capacity factor of 0 is possible but is recommended only in cases where nodes are not joined to the cluster long enough to be useful data owners.

You might need to configure a whole node where the capacity factor is 0 for every cache, user defined caches and internal caches. When defining a zero capacity node, the node won’t hold any data. This is how you declare a zero capacity node:

   <cache-container zero-capacity-node="true" />

new GlobalConfigurationBuilder().zeroCapacityNode(true);

However, note that this will be true for distributed caches only. If you are using replicated caches, the node will still keep a copy of the value. Use only distributed caches to make the best use of this feature.

This is how you configure hashing declaratively, via XML:

   <distributed-cache name="distributedCache" owners="2" segments="100" capacity-factor="2" />

And this is how you can configure it programmatically, in Java:

Configuration c = new ConfigurationBuilder()
   .clustering()
      .cacheMode(CacheMode.DIST_SYNC)
      .hash()
         .numOwners(2)
         .numSegments(100)
         .capacityFactor(2)
   .build();

The hints are configured at transport level:

<transport
    cluster="MyCluster"
    machine="LinuxServer01"
    rack="Rack01"
    site="US-WestCoast" />

Following code snippet depicts how a reference to this service can be obtained and used.

// 1. Obtain a reference to a cache
Cache cache = ...
Address address = cache.getCacheManager().getAddress();

// 2. Create the affinity service
KeyAffinityService keyAffinityService = KeyAffinityServiceFactory.newLocalKeyAffinityService(
      cache,
      new RndKeyGenerator(),
      Executors.newSingleThreadExecutor(),
      100);

// 3. Obtain a key for which the local node is the primary owner
Object localKey = keyAffinityService.getKeyForAddress(address);

// 4. Insert the key in the cache
cache.put(localKey, "yourValue");

The service is started at step 2: after this point it uses the supplied Executor to generate and queue keys. At step 3, we obtain a key from the service, and at step 4 we use it.

KeyAffinityService extends Lifecycle, which allows stopping and (re)starting it:

public interface Lifecycle {
   void start();
   void stop();
}

The service is instantiated through KeyAffinityServiceFactory. All the factory methods have an Executor parameter, that is used for asynchronous key generation (so that it won’t happen in the caller’s thread). It is the user’s responsibility to handle the shutdown of this Executor.

The KeyAffinityService, once started, needs to be explicitly stopped. This stops the background key generation and releases other held resources.

The only situation in which KeyAffinityService stops by itself is when the cache manager with which it was registered is shutdown.

When the cache topology changes (i.e. nodes join or leave the cluster), the ownership of the keys generated by the KeyAffinityService might change. The key affinity service keep tracks of these topology changes and doesn’t return keys that would currently map to a different node, but it won’t do anything about keys generated earlier.

As such, applications should treat KeyAffinityService purely as an optimization, and they should not rely on the location of a generated key for correctness.

In particular, applications should not rely on keys generated by KeyAffinityService for the same address to always be located together. Collocation of keys is only provided by the Grouping API.

By default, the segment of a key is computed using the key’s hashCode(). If you use the grouping API, Red Hat Data Grid will compute the segment of the group and use that as the segment of the key. See the Key Ownership section for more details on how segments are then mapped to nodes.

When the group API is in use, it is important that every node can still compute the owners of every key without contacting other nodes. For this reason, the group cannot be specified manually. The group can either be intrinsic to the entry (generated by the key class) or extrinsic (generated by an external function).

First, you must enable groups. If you are configuring Red Hat Data Grid programmatically, then call:

Configuration c = new ConfigurationBuilder()
   .clustering().hash().groups().enabled()
   .build();

Or, if you are using XML:

<distributed-cache>
   <groups enabled="true"/>
</distributed-cache>

If you have control of the key class (you can alter the class definition, it’s not part of an unmodifiable library), then we recommend using an intrinsic group. The intrinsic group is specified by adding the @Group annotation to a method. Let’s take a look at an example:

class User {
   ...
   String office;
   ...

   public int hashCode() {
      // Defines the hash for the key, normally used to determine location
      ...
   }

   // Override the location by specifying a group
   // All keys in the same group end up with the same owners
   @Group
   public String getOffice() {
      return office;
   }
   }
}

If you don’t have control over the key class, or the determination of the group is an orthogonal concern to the key class, we recommend using an extrinsic group. An extrinsic group is specified by implementing the Grouper interface.

public interface Grouper<T> {
    String computeGroup(T key, String group);

    Class<T> getKeyType();
}

If multiple Grouper classes are configured for the same key type, all of them will be called, receiving the value computed by the previous one. If the key class also has a @Group annotation, the first Grouper will receive the group computed by the annotated method. This allows you even greater control over the group when using an intrinsic group. Let’s take a look at an example Grouper implementation:

public class KXGrouper implements Grouper<String> {

   // The pattern requires a String key, of length 2, where the first character is
   // "k" and the second character is a digit. We take that digit, and perform
   // modular arithmetic on it to assign it to group "0" or group "1".
   private static Pattern kPattern = Pattern.compile("(^k)(<a>\\d</a>)$");

   public String computeGroup(String key, String group) {
      Matcher matcher = kPattern.matcher(key);
      if (matcher.matches()) {
         String g = Integer.parseInt(matcher.group(2)) % 2 + "";
         return g;
      } else {
         return null;
      }
   }

   public Class<String> getKeyType() {
      return String.class;
   }
}

Grouper implementations must be registered explicitly in the cache configuration. If you are configuring Red Hat Data Grid programmatically:

Configuration c = new ConfigurationBuilder()
   .clustering().hash().groups().enabled().addGrouper(new KXGrouper())
   .build();

Or, if you are using XML:

<distributed-cache>
   <groups enabled="true">
      <grouper class="com.acme.KXGrouper" />
   </groups>
</distributed-cache>

AdvancedCache has two group-specific methods:

Both methods iterate over the entire data container and store (if present), so they can be slow when a cache contains lots of small groups.

In this section, we detail how each partition handling strategy behaves in the event of split brain occurring.

Two partitions could start up isolated, and as long as they don’t merge they can read and write inconsistent data. In the future, we will allow custom availability strategies (e.g. check that a certain node is part of the cluster, or check that an external machine is accessible) that could handle that situation as well.

Conflicts are detected by retrieving each of the stored values for a given key. The conflict manager retrieves the value stored from each of the key’s write owners defined by the current consistent hash. The .equals method of the stored values is then used to determine whether all values are equal. If all values are equal then no conflicts exist for the key, otherwise a conflict has occurred. Note that null values are returned if no entry exists on a given node, therefore we deem a conflict to have occurred if both a null and non-null value exists for a given key.

In the event of conflicts arising between one or more replicas of a given CacheEntry, it is necessary for a conflict resolution algorithm to be defined, therefore we provide the EntryMergePolicy interface. This interface consists of a single method, "merge", whose returned CacheEntry is utilised as the "resolved" entry for a given key. When a non-null CacheEntry is returned, this entries value is "put" to all replicas in the cache. However when the merge implementation returns a null value, all replicas associated with the conflicting key are removed from the cache.

The merge method takes two parameters: the "preferredEntry" and "otherEntries". In the context of a partition merge, the preferredEntry is the primary replica of a CacheEntry stored in the partition that contains the most nodes or if partitions are equal the one with the largest topologyId. In the event of overlapping partitions, i.e. a node A is present in the topology of both partitions {A}, {A,B,C}, we pick {A} as the preferred partition as it will have the higher topologId as the other partition’s topology is behind. When a partition merge is not occurring, the "preferredEntry" is simply the primary replica of the CacheEntry. The second parameter, "otherEntries" is simply a list of all other entries associated with the key for which a conflict was detected.

Currently Red Hat Data Grid provides the following implementations of EntryMergePolicy:

It’s also possible to provide custom implementations of the EntryMergePolicy

<distributed-cache name="the-default-cache">
   <partition-handling when-split="ALLOW_READ_WRITES" merge-policy="org.example.CustomMergePolicy"/>
</distributed-cache>

ConfigurationBuilder dcc = new ConfigurationBuilder();
dcc.clustering().partitionHandling()
                    .whenSplit(PartitionHandling.ALLOW_READ_WRITES)
                    .mergePolicy(new CustomMergePolicy());

public class CustomMergePolicy implements EntryMergePolicy<String, String> {

   @Override
   public CacheEntry<String, String> merge(CacheEntry<String, String> preferredEntry, List<CacheEntry<String, String>> otherEntries) {
      // decide which entry should be used

      return the_solved_CacheEntry;
   }

To utilise a custom EntryMergePolicy implementation on the server, it’s necessary for the implementation class(es) to be deployed to the server. This is accomplished by utilising the java service-provider convention and packaging the class files in a jar which has a META-INF/services/org.infinispan.conflict.EntryMergePolicy file containing the fully qualified class name of the EntryMergePolicy implementation.

# list all necessary implementations of EntryMergePolicy with the full qualified name
org.example.CustomMergePolicy

In order for a Custom merge policy to be utilised on the server, you should enable object storage, if your policies semantics require access to the stored Key/Value objects. This is because cache entries in the server may be stored in a marshalled format and the Key/Value objects returned to your policy would be instances of WrappedByteArray. However, if the custom policy only depends on the metadata associated with a cache entry, then object storage is not required and should be avoided (unless needed for other reasons) due to the additional performance cost of marshalling data per request. Finally, object storage is never required if one of the provided merge policies is used.