Hazelcast IMDG Standard Support has expired. Extended support for version 4.1 ends in April 2024. Extended support for version 4.2 ends in September 2024.

We recommend that you try Hazelcast Platform.

In Hazelcast Platform, we’ve combined the in-memory storage of IMDG with the stream processing power of Jet. Find out more in our Platform documentation.

The following topics are a good place to start:

Java Client

The Java client is the most full featured Hazelcast native client. It is offered both with Hazelcast IMDG and Hazelcast IMDG Enterprise. The main idea behind the Java client is to provide the same Hazelcast functionality by proxying each operation through a Hazelcast member. It can access and change distributed data and it can listen to distributed events of an already established Hazelcast cluster from another Java application.

Hundreds or even thousands of clients can be connected to the cluster. By default, there are core count * 20 threads on the server side that handle all the requests, e.g., if the server has 4 cores, there will be 80 threads.

Imagine a trading application where all the trading data are stored and managed in a Hazelcast cluster with tens of members. Swing/Web applications at the traders' desktops can use clients to access and modify the data in the Hazelcast cluster.

Getting Started with Java Client

You do not need to set a license key for your Java clients for which you want to use Hazelcast IMDG Enterprise features. Hazelcast IMDG Enterprise license keys are required only for members.

Simply include the hazelcast.jar dependency in your classpath to start using the Hazelcast Java client. Once included, you can start using this client as if you are using the Hazelcast API. The differences are discussed in the below sections.

If you prefer to use Maven, simply add the hazelcast dependency to your pom.xml, which you may already have done to start using Hazelcast IMDG:

<dependency>
    <groupId>com.hazelcast</groupId>
    <artifactId>hazelcast</artifactId>
    <version>4.2.8</version>
</dependency>

You can find Hazelcast Java client’s code samples here.

Client API

The first step is the configuration. You can configure the Java client declaratively or programmatically. We use the programmatic approach for this section, as shown below.

ClientConfig clientConfig = new ClientConfig();
clientConfig.setClusterName("dev");
clientConfig.getNetworkConfig().addAddress("10.90.0.1", "10.90.0.2:5702");

See the Configuring Java Client section for more information.

The second step is initializing the HazelcastInstance to be connected to the cluster.

HazelcastInstance client = HazelcastClient.newHazelcastClient(clientConfig);

This client interface is your gateway to access all Hazelcast distributed objects.

Let’s create a map and populate it with some data.

IMap<String, Customer> mapCustomers = client.getMap("customers"); //creates the map proxy

mapCustomers.put("1", new Customer("Joe", "Smith"));
mapCustomers.put("2", new Customer("Ali", "Selam"));
mapCustomers.put("3", new Customer("Avi", "Noyan"));

As the final step, if and when you are done with your client, you can shut it down as shown below:

client.shutdown();

The above code line releases all the used resources and closes connections to the cluster.

Java Client Operation Modes

The client has two operation modes because of the distributed nature of the data and cluster.

Smart Client: In the smart mode, the clients connect to each cluster member. Since each data partition uses the well known and consistent hashing algorithm, each client can send an operation to the relevant cluster member, which increases the overall throughput and efficiency. Smart mode is the default mode.

Unisocket Client: For some cases, the clients can be required to connect to a single member instead of to each member in the cluster. Firewalls, security, or some custom networking issues can be the reason for these cases.

In the unisocket client mode, the clients only connect to one of the configured addresses. This single member behaves as a gateway to the other members. For any operation requested from the client, it redirects the request to the relevant member and returns the response back to the client returned from that member.

Using Supported Distributed Data Structures

Most of the Distributed Data Structures are supported by the Java client. When you use clients in other languages, you should check for the exceptions.

As a general rule, you configure these data structures on the server side and access them through a proxy on the client side.

Using Map with Java Client

You can use any Distributed Map object with the client, as shown below.

Imap<Integer, String> map = client.getMap("myMap");

map.put(1, "John");
String value= map.get(1);
map.remove(1);

Locality is ambiguous for the client, so addLocalEntryListener and localKeySet are not supported. See the Distributed Map section for more information.

Using MultiMap with Java Client

A MultiMap usage example is shown below.

MultiMap<Integer, String> multiMap = client.getMultiMap("myMultiMap");

multiMap.put(1,"John");
multiMap.put(1,"Mary");

Collection<String> values = multiMap.get(1);

addLocalEntryListener, localKeySet and getLocalMultiMapStats are not supported because locality is ambiguous for the client. See the Distributed MultiMap section for more information.

Using Queue with Java Client

An example usage is shown below.

IQueue<String> myQueue = client.getQueue("theQueue");
myQueue.offer("John")

getLocalQueueStats is not supported because locality is ambiguous for the client. See the Distributed Queue section for more information.

Using Topic with Java Client

getLocalTopicStats is not supported because locality is ambiguous for the client.

Using Other Supported Distributed Structures

The distributed data structures listed below are also supported by the client. Since their logic is the same in both the member side and client side, you can see their sections as listed below.

Using Client Services

Hazelcast provides the services discussed below for some common functionalities on the client side.

Using Distributed Executor Service

The distributed executor service is for distributed computing. It can be used to execute tasks on the cluster on a designated partition or on all the partitions. It can also be used to process entries. See the Distributed Executor Service section for more information.

IExecutorService executorService = client.getExecutorService("default");

After getting an instance of IExecutorService, you can use the instance as the interface with the one provided on the server side. See the Distributed Computing chapter for detailed usage.

This service is only supported by the Java client.

Listening to Client Connection

If you need to track clients and you want to listen to their connection events, you can use the clientConnected and clientDisconnected methods of the ClientService class. This class must be run on the member side. The following is an example code.

        ClientConfig clientConfig = new ClientConfig();
        //clientConfig.setClusterName("dev");
        clientConfig.getNetworkConfig().addAddress("10.90.0.1", "10.90.0.2:5702");

        HazelcastInstance instance = Hazelcast.newHazelcastInstance();

        final ClientService clientService = instance.getClientService();

        clientService.addClientListener(new ClientListener() {
            @Override
            public void clientConnected(Client client) {
                //Handle client connected event
            }

            @Override
            public void clientDisconnected(Client client) {
                //Handle client disconnected event
            }
        });

        //this will trigger `clientConnected` event
        HazelcastInstance client = HazelcastClient.newHazelcastClient(clientConfig);
        
        final Collection<Client> connectedClients = clientService.getConnectedClients();

        //this will trigger `clientDisconnected` event
        client.shutdown();

Finding the Partition of a Key

You use partition service to find the partition of a key. It returns all partitions. See the example code below.

PartitionService partitionService = client.getPartitionService();

//partition of a key
Partition partition = partitionService.getPartition(key);

//all partitions
Set<Partition> partitions = partitionService.getPartitions();

Handling Lifecycle

Lifecycle handling performs:

  • checking if the client is running

  • shutting down the client gracefully

  • terminating the client ungracefully (forced shutdown)

  • adding/removing lifecycle listeners.

LifecycleService lifecycleService = client.getLifecycleService();

if(lifecycleService.isRunning()){
    //it is running
}

//shutdown client gracefully
lifecycleService.shutdown();

Defining Client Labels

You can define labels in your Java client, similar to the way it can be done for the members. Through the client labels, you can assign special roles for your clients and use these roles to perform some actions specific to those client connections.

You can also group your clients using the client labels. These client groups can be blacklisted in the Hazelcast Management Center so that they can be prevented from connecting to a cluster. See the related section in the Hazelcast Management Center Reference Manual for more information on this topic.

Declaratively, you can define the client labels using the client-labels configuration element. See the below example.

  • XML

  • YAML

<hazelcast-client>
    ...
    <instance-name>barClient</instance-name>
    <client-labels>
        <label>user</label>
        <label>bar</label>
    </client-labels>
    ....
</hazelcast-client>
hazelcast-client:
  instance-name: barClient
  client-labels:
    - user
    - bar

The equivalent programmatic approach is shown below.

ClientConfig clientConfig = new ClientConfig();
clientConfig.setInstanceName("ExampleClientName");
clientConfig.addLabel("user");
clientConfig.addLabel("bar");

HazelcastClient.newHazelcastClient(clientConfig);

See the code sample for the client labels to see them in action.

Client Listeners

You can configure listeners to listen to various event types on the client side. You can configure global events not relating to any distributed object through Client ListenerConfig. You should configure distributed object listeners like map entry listeners or list item listeners through their proxies. See the related sections under each distributed data structure in this Reference Manual.

Client Transactions

Transactional distributed objects are supported on the client side. See the Transactions chapter on how to use them.

Async Start and Reconnect Modes

Java client can be configured to connect to a cluster in an async manner during the client start and reconnecting after a cluster disconnect. Both of these options are configured via ClientConnectionStrategyConfig.

Async client start is configured by setting the configuration element async-start to true. This configuration changes the behavior of HazelcastClient.newHazelcastClient() call. It returns a client instance without waiting to establish a cluster connection. Until the client connects to cluster, it throws HazelcastClientOfflineException on any network dependent operations hence they won’t block. If you want to check or wait the client to complete its cluster connection, you can use the built-in lifecycle listener:

ClientStateListener clientStateListener = new ClientStateListener(clientConfig);
HazelcastInstance client = HazelcastClient.newHazelcastClient(clientConfig);

//Client started but may not be connected to cluster yet.

//check connection status
clientStateListener.isConnected();

//blocks until client completes connect to cluster
if (clientStateListener.awaitConnected()) {
	//connected successfully
} else {
	//client failed to connect to cluster
}

The Java client can also be configured to specify how it reconnects after a cluster disconnection. The following are the options:

  • A client can reject to reconnect to the cluster and trigger the client shutdown process.

  • Client can open a connection to the cluster by blocking all waiting invocations.

  • Client can open a connection to the cluster without blocking the waiting invocations. All invocations receive HazelcastClientOfflineException during the establishment of cluster connection. If cluster connection is failed to connect, then client shutdown is triggered.

See the Java Client Connection Strategy section to learn how to configure these.

Configuring Java Client

You can configure Hazelcast Java Client declaratively (XML), programmatically (API), or using client system properties.

For declarative configuration, the Hazelcast client looks at the following places for the client configuration file:

  • System property: The client first checks if hazelcast.client.config system property is set to a file path, e.g., -Dhazelcast.client.config=C:/myhazelcast.xml.

  • Classpath: If config file is not set as a system property, the client checks the classpath for hazelcast-client.xml file.

If the client does not find any configuration file, it starts with the default configuration (hazelcast-client-default.xml) located in the hazelcast.jar library. Before configuring the client, please try to work with the default configuration to see if it works for you. The default should be just fine for most users. If not, then consider custom configuration for your environment.

If you want to specify your own configuration file to create a Config object, the Hazelcast client supports the following:

  • Config cfg = new XmlClientConfigBuilder(xmlFileName).build();

  • Config cfg = new XmlClientConfigBuilder(inputStream).build();

For programmatic configuration of the Hazelcast Java Client, just instantiate a ClientConfig object and configure the desired aspects. An example is shown below:

ClientConfig clientConfig = new ClientConfig();
clientConfig.setClusterName("dev");
clientConfig.setLoadBalancer(yourLoadBalancer);

Client Network

All network related configuration of Hazelcast Java Client is performed via the network element in the declarative configuration file, or in the class ClientNetworkConfig when using programmatic configuration. Let’s first give the examples for these two approaches. Then we will look at its sub-elements and attributes.

Declarative Configuration:

Here is an example declarative configuration of network for Java Client, which includes all the parent configuration elements.

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <cluster-members>
            <address>127.0.0.1</address>
            <address>127.0.0.2</address>
        </cluster-members>
        <outbound-ports>
            <ports>34600</ports>
            <ports>34700-34710</ports>
        </outbound-ports>
        <smart-routing>true</smart-routing>
        <redo-operation>true</redo-operation>
        <connection-timeout>60000</connection-timeout>
        <socket-options>
            ...
        </socket-options>
        <socket-interceptor enabled="true">
            ...
        </socket-interceptor>

        <ssl enabled="false">
            ...
        </ssl>
        <aws enabled="true" connection-timeout-seconds="11">
            ...
        </aws>
        <gcp enabled="false">
            ...
        </gcp>
        <azure enabled="false">
            ...
        </azure>
        <kubernetes enabled="false">
            ...
        </kubernetes>
        <eureka enabled="false">
            ...
        </eureka>
        <icmp-ping enabled="false">
            ...
        </icmp-ping>
        <hazelcast-cloud enabled="false">
            <discovery-token>EXAMPLE_TOKEN</discovery-token>
        </hazelcast-cloud>
        <discovery-strategies>
            <node-filter class="DummyFilterClass" />
            <discovery-strategy class="DummyDiscoveryStrategy1" enabled="true">
                <properties>
                    <property name="key-string">foo</property>
                    <property name="key-int">123</property>
                    <property name="key-boolean">true</property>
                </properties>
            </discovery-strategy>
        </discovery-strategies>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    cluster-members:
      - 127.0.0.1
      - 127.0.0.2
    outbound-ports:
      - 34600
      - 34700-34710
    smart-routing: true
    redo-operation: true
    connection-timeout: 60000
    socket-options:
      ...
    socket-interceptor:
      ...
    ssl:
      enabled: false
      ...
    aws:
      enabled: true
      connection-timeout-seconds: 11
      ...
    gcp:
      enabled: false
      ...
    azure:
      enabled: false
      ...
    kubernetes:
      enabled: false
      ...
    eureka:
      enabled: false
      ...
    icmp-ping:
      enabled: false
      ...
    hazelcast-cloud:
      enabled: false
      discovery-token: EXAMPLE_TOKEN
    discovery-strategies:
      node-filter:
        class: DummyFilterClass
      discovery-strategies:
        - class: DummyDiscoveryStrategy1
          enabled: true
          properties:
            key-string: foo
            key-int: 123
            key-boolean: true

Programmatic Configuration:

Here is an example of configuring network for Java Client programmatically.

        ClientConfig clientConfig = new ClientConfig();
        clientConfig.getConnectionStrategyConfig().getConnectionRetryConfig().setMaxBackoffMillis(5000);
        ClientNetworkConfig networkConfig = clientConfig.getNetworkConfig();
        networkConfig.addAddress("10.1.1.21", "10.1.1.22:5703")
                .setSmartRouting(true)
                .addOutboundPortDefinition("34700-34710")
                .setRedoOperation(true)
                .setConnectionTimeout(5000);

        AwsConfig clientAwsConfig = new AwsConfig();
        clientAwsConfig.setProperty("access-key", "my-access-key")
                .setProperty("secret-key", "my-secret-key")
                .setProperty("region", "us-west-1")
                .setProperty("host-header", "ec2.amazonaws.com")
                .setProperty("security-group-name", ">hazelcast-sg")
                .setProperty("tag-key", "type")
                .setProperty("tag-value", "hz-members")
                .setProperty("iam-role", "s3access")
                .setEnabled(true);
        clientConfig.getNetworkConfig().setAwsConfig(clientAwsConfig);
        HazelcastInstance client = HazelcastClient.newHazelcastClient(clientConfig);

Configuring Backup Acknowledgment

When an operation with sync backup is sent by a client to the Hazelcast member(s), the acknowledgment of the operation’s backup is sent to the client by the backup replica member(s). This improves the performance of the client operations.

By default, backup acknowledgement to the client is enabled for smart clients (unisocket clients do not support it).

Here is an example of configuring the backup acknowledgement for Java Client declaratively.

  • XML

  • YAML

<hazelcast-client ... >
       <backup-ack-to-client-enabled>false</backup-ack-to-client-enabled>
</hazelcast-client>
hazelcast-client:
  backup-ack-to-client: false

And here is its equivalent programmatical configuration.

clientConfig.setBackupAckToClientEnabled(boolean enabled)

You can also fine tune this feature using the following system properties:

  • hazelcast.client.operation.backup.timeout.millis: If an operation has backups, this property specifies how long (in milliseconds) the invocation waits for acks from the backup replicas. If acks are not received from some of the backups, there will not be any rollback on the other successful replicas. Its default value is 5000 milliseconds.

  • hazelcast.client.operation.fail.on.indeterminate.state: When it is true, if an operation has sync backups and acks are not received from backup replicas in time, or the member which owns primary replica of the target partition leaves the cluster, then the invocation fails. However, even if the invocation fails, there will not be any rollback on other successful replicas. It is default value is false.

Configuring Address List

Address List is the initial list of cluster addresses to which the client will connect. The client uses this list to find an alive member. Although it may be enough to give only one address of a member in the cluster (since all members communicate with each other), it is recommended that you give the addresses for all the members.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <cluster-members>
            <address>10.1.1.21</address>
            <address>10.1.1.22:5703</address>
        </cluster-members>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    cluster-members:
      - 10.1.1.21
      - 10.1.1.22:5703

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
ClientNetworkConfig networkConfig = clientConfig.getNetworkConfig();
networkConfig.addAddress("10.1.1.21", "10.1.1.22:5703");

If the port part is omitted, then 5701, 5702 and 5703 are tried in a random order.

You can provide multiple addresses with ports provided or not, as seen above. The provided list is shuffled and tried in random order. Its default value is localhost.

If you have multiple members on a single machine and you are using unisocket clients, we recommend you to set explicit ports for each member. Then you should provide those ports in your client configuration when you give the member addresses (using the address configuration element or addAddress method as exemplified above). This provides faster connections between clients and members. Otherwise, all the load coming from your clients may go through a single member.

Setting Outbound Ports

You may want to restrict outbound ports to be used by Hazelcast-enabled applications. To fulfill this requirement, you can configure Hazelcast Java client to use only defined outbound ports. The following are example configurations.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <outbound-ports>
            <!-- ports between 34700 and 34710 -->
            <ports>34700-34710</ports>
            <!-- comma separated ports -->
            <ports>34700,34701,34702,34703</ports>
            <ports>34700,34705-34710</ports>
        </outbound-ports>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    outbound-ports:
      - 34700-34710
      - 34700,34701,34702,34703
      - 34700,34705-34710

Programmatic Configuration:

...
NetworkConfig networkConfig = config.getNetworkConfig();
// ports between 34700 and 34710
networkConfig.addOutboundPortDefinition("34700-34710");
// comma separated ports
networkConfig.addOutboundPortDefinition("34700,34701,34702,34703");
networkConfig.addOutboundPort(34705);
...
You can use port ranges and/or comma separated ports.

As shown in the programmatic configuration, you use the method addOutboundPort to add only one port. If you need to add a group of ports, then use the method addOutboundPortDefinition.

In the declarative configuration, the element ports can be used for both single and multiple port definitions.

Setting Smart Routing

Smart routing defines whether the client operation mode is smart or unisocket. See Java Client Operation Modes to learn about these modes.

The following are example configurations.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <smart-routing>true</smart-routing>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    smart-routing: true

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
ClientNetworkConfig networkConfig = clientConfig.getNetworkConfig();
networkConfig.setSmartRouting(true);

Its default value is true (smart client mode).

Note that you need to disable smart routing (false) for the clients which want to use temporary permissions defined in a member. See the Handling Permissions section.

Enabling Redo Operation

It enables/disables redo-able operations as described in Handling Retry-able Operation Failure. The following are the example configurations.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <redo-operation>true</redo-operation>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    redo-operation: true

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
ClientNetworkConfig networkConfig = clientConfig.getNetworkConfig();
networkConfig().setRedoOperation(true);

Its default value is false (disabled).

Setting Connection Timeout

Connection timeout is the timeout value in milliseconds for members to accept client connection requests. The following are the example configurations.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <connection-timeout>5000</connection-timeout>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    connection-timeout: 5000

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
clientConfig.getNetworkConfig().setConnectionTimeout(5000);

Its default value is 5000 milliseconds.

Setting a Socket Interceptor

Hazelcast IMDG Enterprise

Following is a client configuration to set a socket intercepter. Any class implementing com.hazelcast.nio.SocketInterceptor is a socket interceptor.

public interface SocketInterceptor {
    void init(Properties properties);
    void onConnect(Socket connectedSocket) throws IOException;
}

SocketInterceptor has two steps. First, it is initialized by the configured properties. Second, it is informed just after the socket is connected using the onConnect method.

SocketInterceptorConfig socketInterceptorConfig = clientConfig
               .getNetworkConfig().getSocketInterceptorConfig();

MyClientSocketInterceptor myClientSocketInterceptor = new MyClientSocketInterceptor();

socketInterceptorConfig.setEnabled(true);
socketInterceptorConfig.setImplementation(myClientSocketInterceptor);

If you want to configure the socket interceptor with a class name instead of an instance, see the example below.

SocketInterceptorConfig socketInterceptorConfig = clientConfig
            .getNetworkConfig().getSocketInterceptorConfig();

socketInterceptorConfig.setEnabled(true);

//These properties are provided to interceptor during init
socketInterceptorConfig.setProperty("kerberos-host","kerb-host-name");
socketInterceptorConfig.setProperty("kerberos-config-file","kerb.conf");

socketInterceptorConfig.setClassName(MyClientSocketInterceptor.class.getName());
See the Socket Interceptor section for more information.

Configuring Network Socket Options

You can configure the network socket options using SocketOptions. It has the following methods:

  • socketOptions.setKeepAlive(x): Enables/disables the SO_KEEPALIVE socket option. Its default value is true.

  • socketOptions.setTcpNoDelay(x): Enables/disables the TCP_NODELAY socket option. Its default value is true.

  • socketOptions.setReuseAddress(x): Enables/disables the SO_REUSEADDR socket option. Its default value is true.

  • socketOptions.setLingerSeconds(x): Enables/disables SO_LINGER with the specified linger time in seconds. Its default value is 3.

  • socketOptions.setBufferSize(x): Sets the SO_SNDBUF and SO_RCVBUF options to the specified value in KB for this Socket. Its default value is 32.

SocketOptions socketOptions = clientConfig.getNetworkConfig().getSocketOptions();
socketOptions.setBufferSize(32)
             .setKeepAlive(true)
             .setTcpNoDelay(true)
             .setReuseAddress(true)
             .setLingerSeconds(3);

Enabling Client TLS/SSL

Hazelcast IMDG Enterprise

You can use TLS/SSL to secure the connection between the client and the members. If you want TLS/SSL enabled for the client-cluster connection, you should set SSLConfig. Once set, the connection (socket) is established out of an TLS/SSL factory defined either by a factory class name or factory implementation. See the TLS/SSL section.

As explained in the TLS/SSL section, Hazelcast members have keyStores used to identify themselves (to other members) and Hazelcast clients have trustStore used to define which members they can trust. The clients also have their keyStores and members have their trustStores so that the members can know which clients they can trust: see the Mutual Authentication section.

Configuring Hazelcast Cloud

You can connect your Java client to a Hazelcast cluster which is hosted on Cloud. For this, you simply enable Hazelcast Cloud and specify the cluster’s discovery token provided by Hazelcast Cloud while creating the cluster; this allows the Hazelcast cluster to discover your clients. See the following example configurations.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <ssl enabled="true"/>
        <hazelcast-cloud enabled="true">
            <discovery-token>YOUR_TOKEN</discovery-token>
        </hazelcast-cloud>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    ssl:
      enabled: true
    hazelcast-cloud:
      enabled: true
      discovery-token: YOUR_TOKEN

Programmatic Configuration:

ClientConfig config = new ClientConfig();
ClientNetworkConfig networkConfig = config.getNetworkConfig();
networkConfig.getCloudConfig().setDiscoveryToken("TOKEN").setEnabled(true);
networkConfig.setSSLConfig(new SSLConfig().setEnabled(true));
HazelcastInstance client = HazelcastClient.newHazelcastClient(config);

Hazelcast Cloud is disabled for the Java client, by default (enabled attribute is false).

See this Hazelcast Cloud web page for more information.

Since this is a REST based discovery, you need to enable the REST listener service. See the Using the REST Endpoint Groups section on how to enable REST endpoints.

It is advised to enable certificate revocation status JRE-wide, for security reasons. You need to set the following Java system properties to true:

  • com.sun.net.ssl.checkRevocation

  • com.sun.security.enableCRLDP

And you need to set the Java security property as follows:

Security.setProperty("ocsp.enable", "true")

You can find more details on the related security topics here and here.

Configuring Client for AWS

The example declarative and programmatic configurations below show how to configure a Java client for connecting to a Hazelcast cluster in AWS.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <aws enabled="true">
            <use-public-ip>true</use-public-ip>
            <access-key>my-access-key</access-key>
            <secret-key>my-secret-key</secret-key>
            <region>us-west-1</region>
            <host-header>ec2.amazonaws.com</host-header>
            <security-group-name>hazelcast-sg</security-group-name>
            <tag-key>type</tag-key>
            <tag-value>hz-members</tag-value>
        </aws>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    aws:
      enabled: true
      use-public-ip: true
      access-key: my-access-key
      secret-key: my-secret-key
      region: us-west-1
      host-header: ec2.amazonaws.com
      security-group-name: hazelcast-sg
      tag-key: type
      tag-value: hz-members

Programmatic Configuration:

        ClientConfig clientConfig = new ClientConfig();
        AwsConfig clientAwsConfig = new AwsConfig();
        clientAwsConfig.setProperty("access-key", "my-access-key")
                .setProperty("secret-key", "my-secret-key")
                .setProperty("region", "us-west-1")
                .setProperty("host-header", "ec2.amazonaws.com")
                .setProperty("security-group-name", ">hazelcast-sg")
                .setProperty("tag-key", "type")
                .setProperty("tag-value", "hz-members")
                .setProperty("iam-role", "s3access")
                .setEnabled(true);
        clientConfig.getNetworkConfig().setAwsConfig(clientAwsConfig);
        HazelcastInstance client = HazelcastClient.newHazelcastClient(clientConfig);

See the aws element section for the descriptions of the above AWS configuration elements except use-public-ip.

If the use-public-ip element is set to true, the private addresses of cluster members are always converted to public addresses. Also, the client uses public addresses to connect to the members. In order to use private addresses, set the use-public-ip parameter to false. Also note that, when connecting outside from AWS, setting the use-public-ip parameter to false causes the client to not be able to reach the members.

Configuring Client Load Balancer

LoadBalancer allows you to send operations to one of a number of endpoints (Members). Its main purpose is to determine the next Member if queried. It is up to your implementation to use different load balancing policies. You should implement the interface com.hazelcast.client.LoadBalancer for that purpose.

If it is a smart client, only the operations that are not key-based are routed to the endpoint that is returned by the LoadBalancer. If it is not a smart client, LoadBalancer is ignored.

The following are example configurations.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <load-balancer type=“random”/>
    ...
</hazelcast-client>
hazelcast-client:
  load-balancer:
    type: random

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
clientConfig.setLoadBalancer(yourLoadBalancer);

Configuring Client Listeners

You can configure global event listeners using ListenerConfig as shown below.

ClientConfig clientConfig = new ClientConfig();
ListenerConfig listenerConfig = new ListenerConfig(LifecycleListenerImpl);
clientConfig.addListenerConfig(listenerConfig);
ClientConfig clientConfig = new ClientConfig();
ListenerConfig listenerConfig = new ListenerConfig("com.hazelcast.example.MembershipListenerImpl");
clientConfig.addListenerConfig(listenerConfig);

You can add the following types of event listeners:

  • LifecycleListener

  • MembershipListener

  • DistributedObjectListener

Configuring Client Near Cache

The Hazelcast distributed map supports a local Near Cache for remotely stored entries to increase the performance of local read operations. Since the client always requests data from the cluster members, it can be helpful in some use cases to configure a Near Cache on the client side. See the Near Cache section for a detailed explanation of the Near Cache feature and its configuration.

Configuring Client Cluster

Clients should provide a cluster name in order to connect to the cluster. You can configure it using ClientConfig, as shown below.

clientConfig.setClusterName("dev");

Configuring Client Security

In the cases where the security established with Config is not enough and you want your clients connecting securely to the cluster, you can use ClientSecurityConfig. This configuration has a credentials parameter to set the IP address and UID. See the ClientSecurityConfig Javadoc.

Client Serialization Configuration

For the client side serialization, use the Hazelcast configuration. See the Serialization chapter.

Configuring ClassLoader

You can configure a custom classLoader. It is used by the serialization service and to load any class configured in configuration, such as event listeners or ProxyFactories.

Configuring Reliable Topic on the Client Side

Normally when a client uses a Hazelcast data structure, that structure is configured on the member side and the client makes use of that configuration. For the Reliable Topic structure, this is not the case; since it is backed by Ringbuffer, you should configure it on the client side. The class used for this configuration is ClientReliableTopicConfig.

Here is an example programmatic configuration snippet:

        Config config = new Config();
        RingbufferConfig ringbufferConfig = new RingbufferConfig("default");
        ringbufferConfig.setCapacity(10000000)
                .setTimeToLiveSeconds(5);
        config.addRingBufferConfig(ringbufferConfig);

        ClientConfig clientConfig = new ClientConfig();
        ClientReliableTopicConfig topicConfig = new ClientReliableTopicConfig("default");
        topicConfig.setTopicOverloadPolicy( TopicOverloadPolicy.BLOCK )
                            .setReadBatchSize( 10 );
        clientConfig.addReliableTopicConfig(topicConfig);

        HazelcastInstance hz = Hazelcast.newHazelcastInstance(config);
        HazelcastInstance client = HazelcastClient.newHazelcastClient(clientConfig);
        ITopic topic = client.getReliableTopic(topicConfig.getName());

Note that, when you create a Reliable Topic structure on your client, a Ringbuffer (with the same name as the Reliable Topic) is automatically created on the member side, with its default configuration. See the Configuring Ringbuffer section for the defaults. You can edit that configuration according to your needs.

You can configure a Reliable Topic structure on the client side also declaratively. The following is the declarative configuration equivalent to the above example:

  • XML

  • YAML

<hazelcast-client>
    ...
    <ringbuffer name="default">
        <capacity>10000000</capacity>
        <time-to-live-seconds>5</time-to-live-seconds>
    </ringbuffer>
    <reliable-topic name="default">
        <topic-overload-policy>BLOCK</topic-overload-policy>
        <read-batch-size>10</read-batch-size>
    </reliable-topic>
    ...
</hazelcast-client>
hazelcast-client:
  ringbuffer:
    default:
      capacity: 10000000
      time-to-live-seconds: 5
  reliable-topic:
    default:
      topic-overload-policy: BLOCK
      read-batch-size: 10

Java Client Connection Strategy

You can configure the client’s starting mode as async or sync using the configuration element async-start. When it is set to true (async), Hazelcast creates the client without waiting a connection to the cluster. In this case, the client instance throws an exception until it connects to the cluster. If it is false, the client is not created until the cluster is ready to use clients and a connection with the cluster is established. Its default value is false (sync)

You can also configure how the client reconnects to the cluster after a disconnection. This is configured using the configuration element reconnect-mode; it has three options (OFF, ON or ASYNC). The option OFF disables the reconnection. ON enables reconnection in a blocking manner where all the waiting invocations are blocked until a cluster connection is established or failed. The option ASYNC enables reconnection in a non-blocking manner where all the waiting invocations receive a HazelcastClientOfflineException. Its default value is ON.

When you have ASYNC as the reconnect-mode and defined a Near Cache for your client, the client functions without interruptions/downtime by communicating the data from its Near Cache, provided that there is non-expired data in it. See here to learn how you can add a Near Cache to your client.

The example declarative and programmatic configurations below show how to configure a Java client’s starting and reconnecting modes.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <connection-strategy async-start="true" reconnect-mode="ASYNC" />
    ...
</hazelcast-client>
hazelcast-client:
  connection-strategy:
    async-start: true
    reconnect-mode: ASYNC

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
clientConfig.getConnectionStrategyConfig()
            .setAsyncStart(true)
            .setReconnectMode(ClientConnectionStrategyConfig.ReconnectMode.ASYNC);

Configuring Client Connection Retry

When the client is disconnected from the cluster or trying to connect to a one for the first time, it searches for new connections. You can configure the frequency of the connection attempts and client shutdown behavior using ConnectionRetryConfig (programmatical approach)/connection-retry (declarative approach).

Below are the example configurations for each.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client>
    ...
    <connection-strategy async-start="false" reconnect-mode="ON">
        <connection-retry>
            <initial-backoff-millis>1000</initial-backoff-millis>
            <max-backoff-millis>60000</max-backoff-millis>
            <multiplier>2</multiplier>
            <cluster-connect-timeout-millis>50000</cluster-connect-timeout-millis>
            <jitter>0.2</jitter>
        </connection-retry>
    </connection-strategy>
    ...
</hazelcast-client>
hazelcast-client:
  connection-strategy:
    async-start: false
    reconnect-mode: ON
    connection-retry:
      initial-backoff-millis: 1000
      max-backoff-millis: 60000
      multiplier: 2
      cluster-connect-timeout-millis: 50000
      jitter: 0.2

Programmatic Configuration:

ClientConfig config = new ClientConfig();
ClientConnectionStrategyConfig connectionStrategyConfig = config.getConnectionStrategyConfig();
ConnectionRetryConfig connectionRetryConfig = connectionStrategyConfig.getConnectionRetryConfig();
connectionRetryConfig.setInitialBackoffMillis(1000)
                     .setMaxBackoffMillis(60000)
                     .setMultiplier(2)
                     .setClusterConnectTimeoutMillis(50000)
                     .setJitter(0.2);

The following are configuration element descriptions:

  • initial-backoff-millis: Specifies how long to wait (backoff), in milliseconds, after the first failure before retrying. Its default value is 1000 ms.

  • max-backoff-millis: Specifies the upper limit for the backoff in milliseconds. Its default value is 30000 ms.

  • multiplier: Factor to multiply the backoff after a failed retry. Its default value is 1.05.

  • cluster-connect-timeout-millis: Timeout value in milliseconds for the client to give up to connect to the current cluster. Its default value is -1, i.e., infinite. For the default value, client will not stop trying to connect to the target cluster (infinite timeout). If the failover client is used with the default value of this configuration element, the failover client will try to connect alternative clusters after 120000 ms (2 minutes). For any other value, both the client and the failover client will use this as it is.

  • jitter: Specifies by how much to randomize backoffs. Its default value is 0.

A pseudo-code is as follows:

 begin_time = getCurrentTime()
 current_backoff_millis = INITIAL_BACKOFF_MILLIS
 while (TryConnect(connectionTimeout)) != SUCCESS) {
    if (getCurrentTime() - begin_time >= CLUSTER_CONNECT_TIMEOUT_MILLIS) {
         // Give up to connecting to the current cluster and switch to another if exists.
         // For the default values, CLUSTER_CONNECT_TIMEOUT_MILLIS is infinite for the
         // client and equal to the 120000 ms (2 minutes) for the failover client.
    }
    Sleep(current_backoff_millis + UniformRandom(-JITTER * current_backoff_millis, JITTER * current_backoff_millis))
    current_backoff = Min(current_backoff_millis * MULTIPLIER, MAX_BACKOFF_MILLIS)
}

Note that, TryConnect above tries to connect to any member that the client knows, and for each connection we have a connection timeout; see the Setting Connection Timeout section.

Implementing Blue-Green Deployment using Client Filtering

Hazelcast IMDG Enterprise Feature

Hazelcast provides disaster recovery for the client-cluster connections and can use the well-known blue-green mechanism, so that a Java client is automatically diverted to another cluster on demand or when the intended cluster becomes unavailable.

Using the blue-green system, the clients can connect to another cluster automatically when they are blacklisted from their currently connected cluster. See the Hazelcast Management Center Reference Manual for information on blacklisting the clients.

Blue-Green Mechanism

You can make your clients connect to another cluster by blacklisting them in a cluster and using the blue-green mechanism. This is basically having two alive clusters, one of which is active (blue) and the other one is idle (green).

When you blacklist a client in a cluster, the client which is disconnected from the cluster due to this blacklisting, first tries to connect to another member of the same cluster. This is because the client is not aware if this is a blacklisting or a normal disconnection.

The client’s behavior after this disconnection depends on its reconnect-mode. The following are the options when you are using the blue-green mechanism, i.e., you have alternative clusters for your clients to connect:

  • If reconnect-mode is set to ON, the client changes the cluster and blocks the invocations while doing so.

  • If reconnect-mode is set to ASYNC, the client changes the cluster in the background and throws ClientOfflineException while doing so.

  • If reconnect-mode is set to OFF, the client does not change the cluster; it shuts down immediately.

Here it could be the case that the whole cluster is restarted. In this case, the members in the restarted cluster reject the client’s connection request, since the client is trying to connect to the old cluster. So, the client needs to search for a new cluster, if available and according to the blue-green configuration (see the following configuration related sections in this section).

Consider the following notes for the blue-green mechanism (also valid for the disaster recovery mechanism described in the next section):

  • When a client disconnects from a cluster and connects to a new one the InitialMemberEvent and CLIENT_CHANGED_CLUSTER events are fired.

  • When switching clusters, the client reuses its UUID.

  • The client’s listener service re-registers its listeners to the new cluster; the listener service opens a new connection to all members in the current member list and registers the listeners for each connection.

  • The client’s Near Caches and Continuous Query Caches are cleared when the client joins a new cluster successfully.

  • If the new cluster’s partition size is different, the client is rejected by the cluster. The client is not able to connect to a cluster with different partition count.

Disaster Recovery Mechanism

When one of your clusters is gone due to a failure, the connection between your clients and members in that cluster is gone, too. When a client is disconnected because of a failure in the cluster, it first tries to reconnect to the same cluster.

The client’s behavior after this disconnection depends on its reconnect-mode, and it has the same options that are described in the above section (Blue-Green Mechanism).

If you have provided alternative clusters for your clients to connect, the client tries to connect to those alternative clusters (depending on the reconnect-mode).

When a failover starts, i.e., the client is disconnected and was configured to connect to alternative clusters, the current member list is not considered; the client cuts all the connections before attempting to connect to a new cluster and tries the clusters as configured. See the below configuration related sections.

Ordering of Clusters When Clients Try to Connect

The order of the clusters, that the client will try to connect in a blue-green or disaster recovery scenario, is decided by the order of these cluster declarations as given in the client configuration.

Each time the client is disconnected from a cluster and it cannot connect back to the same one, the configured list is iterated over. Count of these iterations before the client decides to shut down is provided using the try-count configuration element. See the following configuration related sections.

We didn’t go over the configuration yet (see the following configuration related sections), but for the sake of explaining the ordering, assume that you have client-config1, client-config2 and client-config3 in the given order as shown below (in your hazelcast-client-failover XML or YAML file). This means you have three alternative clusters.

  • XML

  • YAML

<hazelcast-client-failover>
    <try-count>4</try-count>
    <clients>
        <client>client-config1.xml</client>
        <client>client-config2.xml</client>
        <client>client-config3.xml</client>
    </clients>
</hazelcast-client-failover>
hazelcast-client-failover:
  try-count: 4
  clients:
    - client-config1.yaml
    - client-config2.yaml
    - client-config3.yaml

And let’s say the client is disconnected from the cluster whose configuration is given by client-config2.xml. Then, the client tries to connect to the next cluster in this list, whose configuration is given by client-config3.xml. When the end of the list is reached, which is so in this example, and the client could not connect to client-config3, then try-count is incremented and the client continues to try to connect starting with client-config1.

This iteration continues until the client connects to a cluster or try-count is reached to the configured value. When the iteration reaches this value and the client still could not connect to a cluster, it shuts down. Note that, if try-count was set to 1 in the above example, and the client could not connect to client-config3, it would shut down since it already tried once to connect to an alternative cluster.

The following sections describe how you can configure the Java client for blue-green and disaster recovery scenarios.

Configuring Using CNAME

Using CNAME, you can change the hostname resolutions and use them dynamically. Let’s describe the configuration with examples.

Assume that you have two clusters, Cluster A and Cluster B, and two Java clients.

First configure the Cluster A members as shown below:

  • XML

  • YAML

<hazelcast>
    ...
    <network>
        <join>
            <tcp-ip enabled="true">
                <member>clusterA.member1</member>
                <member>clusterA.member2</member>
            </tcp-ip>
        </join>
    </network>
    ...
</hazelcast>
hazelcast:
  network:
    join:
      tcp-ip:
        enabled: true
        members: clusterA.member1,clusterA.member2

Then, configure the Cluster B members as shown below.

  • XML

  • YAML

<hazelcast>
    ...
    <network>
        <join>
            <tcp-ip enabled="true">
                <member>clusterB.member1</member>
                <member>clusterB.member2</member>
            </tcp-ip>
        </join>
    </network>
    ...
</hazelcast>
hazelcast:
  network:
    join:
      tcp-ip:
        enabled: true
        members: clusterB.member1,clusterB.member2

Configure the two clients as shown below.

  • Client 1 XML

  • YAML

<hazelcast-client>
    ...
    <cluster-name>cluster-a</cluster-name>
    <network>
        <cluster-members>
            <address>production1.myproject</address>
            <address>production2.myproject</address>
        </cluster-members>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  cluster-name: cluster-a
  network:
    cluster-members:
      - production1.myproject
      - production2.myproject
  • Client 2 XML

  • YAML

<hazelcast-client>
    ...
    <cluster-name>cluster-b</cluster-name>
    <network>
        <cluster-members>
            <address>production1.myproject</address>
            <address>production2.myproject</address>
        </cluster-members>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  cluster-name: cluster-b
  network:
    cluster-members:
      - production1.myproject
      - production2.myproject

Assuming that the client configuration file names of the above example clients are hazelcast-client-c1.xml/yaml and hazelcast-client-c1.xml/yaml, you should configure the client failover for a blue-green deployment scenario as follows:

  • XML

  • YAML

<hazelcast-client-failover>
    <try-count>4</try-count>
    <clients>
        <client>hazelcast-client-c1.xml</client>
        <client>hazelcast-client-c2.xml</client>
    </clients>
</hazelcast-client-failover>
hazelcast-client-failover:
  try-count: 4
  clients:
    - hazelcast-client-c1.yaml
    - hazelcast-client-c2.yaml
You can find the complete Hazelcast client failover example configuration file (hazelcast-client-failover-full-example) both in XML and YAML formats including the descriptions of elements and attributes, in the /bin folder of your Hazelcast download directory.

You should also configure your clients to forget DNS lookups using the networkaddress.cache.ttl JVM parameter.

Configure the addresses in your clients' configuration to resolve to hostnames of Cluster A via CNAME so that the clients will connect to Cluster A when it starts:

production1.myprojectclusterA.member1

production2.myprojectclusterA.member2

When you want the clients to switch to the other cluster, change the mapping as follows:

production1.myprojectclusterB.member1

production2.myprojectclusterB.member2

Wait for the time you configured using the networkaddress.cache.ttl JVM parameter for the client JVM to forget the old mapping.

Blacklist the clients in Cluster A using the Hazelcast Management Center.

Configuring Without CNAME

Let’s first give example configurations and describe the configuration elements.

Declarative Configuration:

  • XML

  • YAML

<hazelcast-client-failover>
    <try-count>4</try-count>
    <clients>
        <client>hazelcast-client-c1.xml</client>
        <client>hazelcast-client-c2.xml</client>
    </clients>
</hazelcast-client-failover>
hazelcast-client-failover:
  try-count: 4
  clients:
    - hazelcast-client-c1.yaml
    - hazelcast-client-c2.yaml

Programmatic Configuration:

ClientConfig clientConfig = new ClientConfig();
clientConfig.setClusterName("cluster-a");
ClientNetworkConfig networkConfig = clientConfig.getNetworkConfig();
networkConfig.addAddress("10.216.1.18", "10.216.1.19");

ClientConfig clientConfig2 = new ClientConfig();
clientConfig2.setClusterName("cluster-b");
ClientNetworkConfig networkConfig2 = clientConfig2.getNetworkConfig();
networkConfig2.addAddress( "10.214.2.10", "10.214.2.11");

ClientFailoverConfig clientFailoverConfig = new ClientFailoverConfig();
clientFailoverConfig.addClientConfig(clientConfig).addClientConfig(clientConfig2).setTryCount(10)
HazelcastInstance client = HazelcastClient.newHazelcastFailoverClient(clientFailoverConfig);

The following are the descriptions for the configuration elements:

  • try-count: Count of connection retries by the client to the alternative clusters. When this value is reached and the client still could not connect to a cluster, the client shuts down. Note that this value applies to the alternative clusters whose configurations are provided with the client element. For the above example, two alternative clusters are given with the try-count set as 4. This means the number of connection attempts is 4 x 2 = 8.

  • client: Path to the client configuration that corresponds to an alternative cluster that the client will try to connect.

The client configurations must be exactly the same except the following configuration options:

  • SecurityConfig

  • NetworkConfig.Addresses

  • NetworkConfig.SocketInterceptorConfig

  • NetworkConfig.SSLConfig

  • NetworkConfig.AwsConfig

  • NetworkConfig.GcpConfig

  • NetworkConfig.AzureConfig

  • NetworkConfig.KubernetesConfig

  • NetworkConfig.EurekaConfig

  • NetworkConfig.CloudConfig

  • NetworkConfig.DiscoveryConfig

You can also configure it within the Spring context, as shown below:

<beans>
    <hz:client-failover id="blueGreenClient" try-count="5">
        <hz:client>
            <hz:cluster-name name="dev"/>
            <hz:network>
                <hz:member>127.0.0.1:5700</hz:member>
                <hz:member>127.0.0.1:5701</hz:member>
            </hz:network>
        </hz:client>

        <hz:client>
            <hz:cluster-name name="alternativeClusterName"/>
            <hz:network>
                <hz:member>127.0.0.1:5702</hz:member>
                <hz:member>127.0.0.1:5703</hz:member>
            </hz:network>
        </hz:client>

    </hz:client-failover>
</beans>

Handling Failures

There are three main failure cases and configurations you can perform to achieve proper behavior.

Handling Client Connection Failure:

While the client is trying to connect initially to one of the members in the ClientNetworkConfig.addressList, all the members might be not available. Instead of giving up, throwing an exception and stopping the client, the client retries to connect as configured which is described in the Configuring Client Connection Retry section.

The client executes each operation through the already established connection to the cluster. If this connection(s) disconnects or drops, the client tries to reconnect as configured.

Handling Unresponsive Connections

It is possible for a connection to be open, but unresponsive, due to several scenarios like JVM pauses or TCP half-open sockets. If this occurs, a client which already has an open connection to that member will not be able to tell that the member is unresponsive because the connection is still open and no errors are received from the member side. Client operations will stay in-flight, potentially indefinitely, in this scenario. To avoid this, Hazelcast provides Client Failure Detectors to mark members which appear to be unresponsive or down. Clients will stop sending requests to members which are marked by the failure detector. Depending upon which operation mode the client is configured to use, it may also attempt to send requests via other members it is aware of.

Handling Retry-able Operation Failure:

While sending the requests to related members, operations can fail due to various reasons. Read-only operations are retried by default. If you want to enable retry for the other operations, you can set the redoOperation to true. See the Enabling Redo Operation section.

You can set a timeout for retrying the operations sent to a member. This can be provided by using the property hazelcast.client.invocation.timeout.seconds in ClientProperties. The client retries an operation within this given period if it is a read-only operation or you enabled the redoOperation as stated in the above paragraph. This timeout value is relevant only when there is a failure of the following types:

  • Member throws an exception

  • Connection between the client and member is closed

  • Client Failure Detectors detect a member is unresponsive

See the Client System Properties section for the description of the hazelcast.client.invocation.timeout.seconds property.

When any failure happens between a client and member (such as an exception on the member side or connection issues), an operation is retried if:

  • it is certain that it has not run on the member yet

  • or if it is idempotent such as a read-only operation, i.e., retrying does not have a side effect.

If it is not certain whether the operation has run on the member, then the non-idempotent operations are not retried. However, as explained in the first paragraph of this section, you can force all client operations to be retried (redoOperation) when there is a failure between the client and member. But in this case, you should know that some operations may run multiple times causing conflicts. For example, assume that your client sent a queue.offer operation to the member and then the connection is lost. Since there will be no respond for this operation, you will not know whether it has run on the member or not. If you enabled redoOperation, that queue.offer operation may rerun and this causes the same objects to be offered twice in the member’s queue.

Java Client Failure Detectors

The client failure detectors are responsible to determine if a member in the cluster is unreachable or crashed. The most important problem in the failure detection is to distinguish whether a member is still alive but slow, or has crashed. But according to the famous FLP result, it is impossible to distinguish a crashed member from a slow one in an asynchronous system. A workaround to this limitation is to use unreliable failure detectors. An unreliable failure detector allows a member to suspect that others have failed, usually based on liveness criteria but it can make mistakes to a certain degree.

Hazelcast Java client has two built-in failure detectors: Deadline Failure Detector and Ping Failure Detector. These client failure detectors work independently from the member failure detectors, e.g., you do not need to enable the member failure detectors to benefit from the client ones.

Client Deadline Failure Detector

Deadline Failure Detector uses an absolute timeout for missing/lost heartbeats. After timeout, a member is considered as crashed/unavailable and marked as suspected.

Deadline Failure Detector has two configuration properties:

  • hazelcast.client.heartbeat.interval: This is the interval at which client sends heartbeat messages to members.

  • hazelcast.client.heartbeat.timeout: This is the timeout which defines when a cluster member is suspected, because it has not sent any response back to client requests.

The value of hazelcast.client.heartbeat.interval should be smaller than that of hazelcast.client.heartbeat.timeout. In addition, the value of system property hazelcast.client.max.no.heartbeat.seconds, which is set on the member side, should be larger than that of hazelcast.client.heartbeat.interval.

The following is a declarative example showing how you can configure the Deadline Failure Detector for your client (in the client’s configuration XML file, e.g., hazelcast-client.xml):

  • XML

  • YAML

<hazelcast-client>
    ...
    <properties>
        <property name="hazelcast.client.heartbeat.timeout">60000</property>
        <property name="hazelcast.client.heartbeat.interval">5000</property>
    </properties>
    ...
</hazelcast-client>
hazelcast-client:
  properties
    hazelcast.client.heartbeat.timeout: 60000
    hazelcast.client.heartbeat.interval: 5000

And, the following is the equivalent programmatic configuration:

ClientConfig config = ...;
config.setProperty("hazelcast.client.heartbeat.timeout", "60000");
config.setProperty("hazelcast.client.heartbeat.interval", "5000");
[...]

Client Ping Failure Detector

In addition to the Deadline Failure Detector, the Ping Failure Detector may be configured on your client. Please note that this detector is disabled by default. The Ping Failure Detector operates at Layer 3 of the OSI protocol and provides much quicker and more deterministic detection of hardware and other lower level events. When the JVM process has enough permissions to create RAW sockets, the implementation chooses to rely on ICMP Echo requests. This is preferred.

If there are not enough permissions, it can be configured to fallback on attempting a TCP Echo on port 7. In the latter case, both a successful connection or an explicit rejection is treated as "Host is Reachable". Or, it can be forced to use only RAW sockets. This is not preferred as each call creates a heavy weight socket and moreover the Echo service is typically disabled.

For the Ping Failure Detector to rely only on the ICMP Echo requests, the following criteria need to be met:

  • Supported OS: as of Java 1.8 only Linux/Unix environments are supported.

  • The Java executable must have the cap_net_raw capability.

  • The file ld.conf must be edited to overcome the rejection by the dynamic linker when loading libs from untrusted paths.

  • ICMP Echo Requests must not be blocked by the receiving hosts.

The details of these requirements are explained in the Requirements section of Hazelcast members' Ping Failure Detector.

If any of the above criteria isn’t met, then isReachable will always fallback on TCP Echo attempts on port 7.

An example declarative configuration to use the Ping Failure Detector is as follows (in the client’s configuration XML file, e.g., hazelcast-client.xml):

  • XML

  • YAML

<hazelcast-client>
    ...
    <network>
        <icmp-ping enabled="true">
            <timeout-milliseconds>1000</timeout-milliseconds>
            <interval-milliseconds>1000</interval-milliseconds>
            <ttl>255<ttl>
            <echo-fail-fast-on-startup>false</echo-fail-fast-on-startup>
            <max-attempts>2</max-attempts>
        </icmp-ping>
    </network>
    ...
</hazelcast-client>
hazelcast-client:
  network:
    icmp-ping:
      enabled: false
      timeout-milliseconds: 1000
      interval-milliseconds: 1000
      ttl: 255
      echo-fail-fast-on-startup: false
      max-attempts: 2

And, the equivalent programmatic configuration:

ClientConfig config = ...;

ClientNetworkConfig networkConfig = clientConfig.getNetworkConfig();
ClientIcmpPingConfig clientIcmpPingConfig = networkConfig.getClientIcmpPingConfig();
clientIcmpPingConfig.setIntervalMilliseconds(1000)
        .setTimeoutMilliseconds(1000)
        .setTtl(255)
        .setMaxAttempts(2)
        .setEchoFailFastOnStartup(false)
        .setEnabled(true);

The following are the descriptions of configuration elements and attributes:

  • enabled: Enables the legacy ICMP detection mode, works cooperatively with the existing failure detector and only kicks-in after a pre-defined period has passed with no heartbeats from a member. Its default value is false.

  • timeout-milliseconds: Number of milliseconds until a ping attempt is considered failed if there was no reply. Its default value is 1000 milliseconds.

  • max-attempts: Maximum number of ping attempts before the member gets suspected by the detector. Its default value is 3.

  • interval-milliseconds: Interval, in milliseconds, between each ping attempt. 1000ms (1 sec) is also the minimum interval allowed. Its default value is 1000 milliseconds.

  • ttl: Maximum number of hops the packets should go through. Its default value is 255. You can set to 0 to use your system’s default TTL.

In the above example configuration, the Ping Failure Detector attempts 2 pings, one every second, and waits up to 1 second for each to complete. If there is no successful ping after 2 seconds, the member gets suspected.

To enforce the Requirements, the property echo-fail-fast-on-startup can also be set to true, in which case Hazelcast fails to start if any of the requirements isn’t met.

Unlike the Hazelcast members, Ping Failure Detector works always in parallel with Deadline Failure Detector on the clients. Below is a summary table of all possible configuration combinations of the Ping Failure Detector.

ICMP Fail-Fast Description Linux Windows macOS

true

false

Parallel ping detector, works in parallel with the configured failure detector. Checks periodically if members are live (OSI Layer 3) and suspects them immediately, regardless of the other detectors.

Supported ICMP Echo if available - Falls back on TCP Echo on port 7

Supported TCP Echo on port 7

Supported ICMP Echo if available - Falls back on TCP Echo on port 7

true

true

Parallel ping detector, works in parallel with the configured failure detector. Checks periodically if members are live (OSI Layer 3) and suspects them immediately, regardless of the other detectors.

Supported - Requires OS Configuration Enforcing ICMP Echo if available - No start up if not available

Not Supported

Not Supported - Requires root privileges

Client System Properties

There are some advanced client configuration properties to tune some aspects of Hazelcast Client. You can set them as property name and value pairs through declarative configuration, programmatic configuration, or JVM system property. See the System Properties appendix to learn how to set these properties.

When you want to reconfigure a system property, you need to restart the clients for which the property is modified.

The table below lists the client configuration properties with their descriptions.

Table 1. Client System Properties
Property Name Default Value Type Description

hazelcast.client.cloud.discovery.token

long

Token to use when discovering the cluster via Hazelcast Cloud.

hazelcast.client.concurrent.window.ms

100

int

Property needed for concurrency detection so that write through and dynamic response handling can be done correctly. This property sets the window for a concurrency detection (duration when it signals that a concurrency has been detected), even if there are no further updates in that window. Normally in a concurrent system the windows keeps sliding forward so it always remains concurrent. Setting it too high effectively disables the optimization because once concurrency has been detected it will keep that way. Setting it too low could lead to suboptimal performance because the system will try write through and other optimizations even though the system is concurrent.

hazelcast.discovery.enabled

false

bool

Enables/disables the Discovery SPI lookup over the old native implementations. See Discovery SPI for more information.

hazelcast.discovery.public.ip.enabled

false

bool

Enables the discovery joiner to use public IPs from DiscoveredNode. See Discovery SPI for more information.

hazelcast.client.event.queue.capacity

1000000

int

Default value of the capacity of executor that handles the incoming event packets.

hazelcast.client.event.thread.count

5

int

Thread count for handling the incoming event packets.

hazelcast.client.heartbeat.interval

5000

int

Frequency of the heartbeat messages sent by the clients to members.

hazelcast.client.heartbeat.timeout

60000

int

Timeout for the heartbeat messages sent by the client to members. If no messages pass between the client and member within the given time via this property in milliseconds, the connection will be closed.

hazelcast.client.invocation.backoff.timeout.millis

-1

int

Controls the maximum timeout, in milliseconds, to wait for an invocation space to be available. If an invocation cannot be made because there are too many pending invocations, then an exponential backoff is done to give the system time to deal with the backlog of invocations. This property controls how long an invocation is allowed to wait before getting a HazelcastOverloadException. When set to -1 then HazelcastOverloadException is thrown immediately without any waiting.

hazelcast.client.invocation.retry.pause.millis

1000

int

Pause time between each retry cycle of an invocation in milliseconds.

hazelcast.client.invocation.timeout.seconds

120

int

Period, in seconds, to give up the invocation when a member in the member list is not reachable.

hazelcast.client.io.balancer.interval.seconds

20

int

Interval in seconds between each IOBalancer execution. By default Hazelcast uses 3 threads to read data from TCP connections and 3 threads to write data to connections. IOBalancer detects and fixes the fluctuations when these threads are not utilized equally. The shorter intervals catch I/O imbalances faster, but they cause higher overhead. A value smaller than 1 disables the balancer.

hazelcast.client.io.input.thread.count

-1

int

Controls the number of I/O input threads. Defaults to -1, i.e., the system decides. If the client is a smart client, it defaults to 3, otherwise it defaults to 1.

hazelcast.client.io.output.thread.count

-1

int

Controls the number of I/O output threads. Defaults to -1, i.e., the system decides. If the client is a smart client, it defaults to 3, otherwise it defaults to 1.

hazelcast.client.io.write.through

true

bool

Optimization that allows sending of packets over the network to be done on the calling thread if the conditions are right. This can reduce the latency and increase the performance for low threaded environments.

hazelcast.client.max.concurrent.invocations

Integer.MAX_VALUE

int

Maximum allowed number of concurrent invocations. You can apply a constraint on the number of concurrent invocations in order to prevent the system from overloading. If the maximum number of concurrent invocations is exceeded and a new invocation comes in, Hazelcast throws HazelcastOverloadException.

hazelcast.client.metrics.collection.frequency

5

int

Frequency, in seconds, of the metrics collection cycle. Note that the preferred way for controlling this setting is Metrics Configuration.

hazelcast.client.metrics.debug.enabled

false

bool

Enables collecting debug metrics if set to true, disables it otherwise. Note that this is meant to be enabled only if diagnostics is enabled, since currently only diagnostics consumes the debug metrics.

hazelcast.client.metrics.enabled

true

bool

Enables the metrics collection if set to true, disables it otherwise. Note that the preferred way for controlling this setting is Metrics Configuration.

hazelcast.client.metrics.jmx.enabled

true

bool

Enables exposing the collected metrics over JMX if set to true, disables it otherwise. Note that the preferred way for controlling this setting is Metrics Configuration.

hazelcast.client.operation.backup.timeout.millis

5000

int

If an operation has backups, this property specifies how long the invocation will wait for acks from the backup replicas. If acks are not received from some backups, there will not be any rollback on other successful replicas.

hazelcast.client.operation.fail.on.indeterminate.state

false

bool

When this configuration is enabled, if an operation has sync backups and acks are not received from backup replicas in time, or the member which owns primary replica of the target partition leaves the cluster, then the invocation fails with IndeterminateOperationStateException. However, even if the invocation fails, there will not be any rollback on other successful replicas.

hazelcast.client.response.thread.count

2

int

Number of the response threads. By default, there are two response threads; this gives stable and good performance. If set to 0, the response threads are bypassed and the response handling is done on the I/O threads. Under certain conditions this can give a higher throughput, but setting to 0 should be regarded as an experimental feature. If set to 0, the IO_OUTPUT_THREAD_COUNT is really going to matter because the inbound thread will have more work to do. By default when TLS is not enabled, there is just one inbound thread.

hazelcast.client.response.thread.dynamic

true

bool

Enables dynamic switching between processing the responses on the I/O threads and offloading the response threads. Under certain conditions (single threaded clients) processing on the I/O thread can increase the performance because useless handover to the response thread is removed. Also the response thread is not created until it is needed. Especially for ephemeral clients, reducing the threads can lead to increased performance and reduced memory usage.

hazelcast.client.shuffle.member.list

true

string

The client shuffles the given member list to prevent all the clients to connect to the same member when this property is true. When it is set to false, the client tries to connect to the members in the given order.

hazelcast.client.statistics.enabled

false

bool

If set to true, it enables collecting the client statistics and sending them to the cluster. When it is true you can monitor the clients that are connected to your Hazelcast cluster, using Hazelcast Management Center. See the Monitoring Clients section^ in the Hazelcast Management Center Reference Manual for more information.

hazelcast.client.statistics.period.seconds

3

int

Period in seconds the client statistics are collected and sent to the cluster. See the Monitoring Clients section in the Hazelcast Management Center Reference Manual for more information on the client statistics.

Using High-Density Memory Store with Java Client

Hazelcast IMDG Enterprise HD

If you have Hazelcast IMDG Enterprise HD, your Hazelcast Java client’s Near Cache can benefit from the High-Density Memory Store.

Let’s recall the Java client’s Near Cache configuration (see the Configuring Client Near Cache section) without High-Density Memory Store:

<hazelcast-client>
    ...
    <near-cache name="MENU">
        <eviction size="2000" eviction-policy="LFU"/>
        <time-to-live-seconds>0</time-to-live-seconds>
        <max-idle-seconds>0</max-idle-seconds>
        <invalidate-on-change>true</invalidate-on-change>
        <in-memory-format>OBJECT</in-memory-format>
    </near-cache>
    ...
</hazelcast-client>

You can configure this Near Cache to use Hazelcast’s High-Density Memory Store by setting the in-memory format to NATIVE. See the following configuration example:

<hazelcast-client>
    ...
    <near-cache>
        <eviction size="1000" max-size-policy="ENTRY_COUNT" eviction-policy="LFU"/>
        <time-to-live-seconds>0</time-to-live-seconds>
        <max-idle-seconds>0</max-idle-seconds>
        <invalidate-on-change>true</invalidate-on-change>
        <in-memory-format>NATIVE</in-memory-format>
    </near-cache>
</hazelcast-client>

The <eviction> element has the following attributes:

  • size: Maximum size (entry count) of the Near Cache.

  • max-size-policy: Maximum size policy for eviction of the Near Cache. Available values are as follows:

    • ENTRY_COUNT: Maximum entry count per member.

    • USED_NATIVE_MEMORY_SIZE: Maximum used native memory size in megabytes.

    • USED_NATIVE_MEMORY_PERCENTAGE: Maximum used native memory percentage.

    • FREE_NATIVE_MEMORY_SIZE: Minimum free native memory size to trigger cleanup.

    • FREE_NATIVE_MEMORY_PERCENTAGE: Minimum free native memory percentage to trigger cleanup.

  • eviction-policy: Eviction policy configuration. Its default values is NONE. Available values are as follows:

    • NONE: No items are evicted and the size property is ignored. You still can combine it with time-to-live-seconds.

    • LRU: Least Recently Used.

    • LFU: Least Frequently Used.

Keep in mind that you should have already enabled the High-Density Memory Store usage for your client, using the <native-memory> element in the client’s configuration.

See the High-Density Memory Store section for more information on Hazelcast’s High-Density Memory Store feature.