Hazelcast Component

Available as of Camel version 2.7

The hazelcast: component allows you to work with the Hazelcast distributed data grid / cache. Hazelcast is a in memory data grid, entirely written in Java (single jar). It offers a great palette of different data stores like map, multi map (same key, n values), queue, list and atomic number. The main reason to use Hazelcast is its simple cluster support. If you have enabled multicast on your network you can run a cluster with hundred nodes with no extra configuration. Hazelcast can simply configured to add additional features like n copies between nodes (default is 1), cache persistence, network configuration (if needed), near cache, enviction and so on. For more information consult the Hazelcast documentation on http://www.hazelcast.com/docs.jsp.

Maven users will need to add the following dependency to their pom.xml for this component:

<dependency>
    <groupId>org.apache.camel</groupId>
    <artifactId>camel-hazelcast</artifactId>
    <version>x.x.x</version>
    <!-- use the same version as your Camel core version -->
</dependency>

URI format

hazelcast:[ map | multimap | queue | topic | seda | set | atomicvalue | instance | list | ringbuffer]:cachename[?options]
  • Topic support is available as of Camel 2.15. *

  • RingBuffer support is available as of Camel 2.16. *

Options

The Hazelcast component supports 3 options which are listed below.

Name Description Default Type

hazelcastInstance (advanced)

The hazelcast instance reference which can be used for hazelcast endpoint. If you don’t specify the instance reference camel use the default hazelcast instance from the camel-hazelcast instance.

HazelcastInstance

hazelcastMode (advanced)

The hazelcast mode reference which kind of instance should be used. If you don’t specify the mode then the node mode will be the default.

node

String

resolveProperty Placeholders (advanced)

Whether the component should resolve property placeholders on itself when starting. Only properties which are of String type can use property placeholders.

true

boolean

The Hazelcast endpoint is configured using URI syntax:

hazelcast:command:cacheName

with the following path and query parameters:

Path Parameters (2 parameters):

Name Description Default Type

command

Required What operation to perform.

HazelcastCommand

cacheName

Required The name of the cache

String

Query Parameters (12 parameters):

Name Description Default Type

defaultOperation (common)

To specify a default operation to use if no operation header has been provided.

String

hazelcastInstance (common)

The hazelcast instance reference which can be used for hazelcast endpoint.

HazelcastInstance

hazelcastInstanceName (common)

The hazelcast instance reference name which can be used for hazelcast endpoint. If you don’t specify the instance reference camel use the default hazelcast instance from the camel-hazelcast instance.

String

reliable (common)

Define if the endpoint will use a reliable Topic struct or not.

false

boolean

bridgeErrorHandler (consumer)

Allows for bridging the consumer to the Camel routing Error Handler which mean any exceptions occurred while the consumer is trying to pickup incoming messages or the likes will now be processed as a message and handled by the routing Error Handler. By default the consumer will use the org.apache.camel.spi.ExceptionHandler to deal with exceptions that will be logged at WARN or ERROR level and ignored.

false

boolean

exceptionHandler (consumer)

To let the consumer use a custom ExceptionHandler. Notice if the option bridgeErrorHandler is enabled then this options is not in use. By default the consumer will deal with exceptions that will be logged at WARN or ERROR level and ignored.

ExceptionHandler

exchangePattern (consumer)

Sets the exchange pattern when the consumer creates an exchange.

ExchangePattern

synchronous (advanced)

Sets whether synchronous processing should be strictly used or Camel is allowed to use asynchronous processing (if supported).

false

boolean

concurrentConsumers (seda)

To use concurrent consumers polling from the SEDA queue.

1

int

pollTimeout (seda)

The timeout used when consuming from the SEDA queue. When a timeout occurs the consumer can check whether it is allowed to continue running. Setting a lower value allows the consumer to react more quickly upon shutdown.

1000

int

transacted (seda)

If set to true then the consumer runs in transaction mode where the messages in the seda queue will only be removed if the transaction commits which happens when the processing is complete.

false

boolean

transferExchange (seda)

If set to true the whole Exchange will be transfered. If header or body contains not serializable objects they will be skipped.

false

boolean

Sections

  1. Usage of #map

  2. Usage of #multimap

  3. Usage of #queue

  4. Usage of #topic

  5. Usage of #list

  6. Usage of #seda

  7. Usage of atomic number

  8. Usage of #cluster support (instance)

  9. Usage of #replicatedmap 

  10. Usage of #ringbuffer 

Usage of Map

Map cache producer - to("hazelcast:map:foo")

If you want to store a value in a map you can use the map cache producer.

The map cache producer provides follow operations specified by CamelHazelcastOperationType header:

  • put

  • putIfAbsent

  • get

  • getAll

  • keySet

  • containsKey

  • containsValue

  • delete

  • update

  • query

  • clear

  • evict

  • evictAll

All operations are provide the inside the "hazelcast.operation.type" header variable. In Java DSL you can use the constants from org.apache.camel.component.hazelcast.HazelcastConstants.

Header Variables for the request message:

Name Type Description

CamelHazelcastOperationType

String

as already described.

CamelHazelcastObjectId

String

the object id to store / find your object inside the cache (not needed for the query operation)

put and putIfAbsent operations provide an eviction mechanism:

Name Type Description

CamelHazelcastObjectTtlValue

Integer

value of TTL.

CamelHazelcastObjectTtlUnit

java.util.concurrent.TimeUnit

value of time unit ( DAYS / HOURS / MINUTES / …​.

You can call the samples with:

template.sendBodyAndHeader("direct:[put|get|update|delete|query|evict]", "my-foo", HazelcastConstants.OBJECT_ID, "4711");

Sample for put:

Java DSL:

from("direct:put")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PUT_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:put" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>put</constant>
    </setHeader>
    <to uri="hazelcast:map:foo" />
</route>

Sample for put with eviction:

Java DSL:

from("direct:put")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PUT_OPERATION))
.setHeader(HazelcastConstants.TTL_VALUE, constant(Long.valueOf(1)))
.setHeader(HazelcastConstants.TTL_UNIT, constant(TimeUnit.MINUTES))
.toF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:put" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>put</constant>
    </setHeader>
    <setHeader headerName="HazelcastConstants.TTL_VALUE">
        <simple resultType="java.lang.Long">1</simple>
    </setHeader>
    <setHeader headerName="HazelcastConstants.TTL_UNIT">
        <simple resultType="java.util.concurrent.TimeUnit">TimeUnit.MINUTES</simple>
    </setHeader>
    <to uri="hazelcast:map:foo" />
</route>

Sample for get:

Java DSL:

from("direct:get")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.GET_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX)
.to("seda:out");

Spring DSL:

<route>
    <from uri="direct:get" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>get</constant>
    </setHeader>
    <to uri="hazelcast:map:foo" />
    <to uri="seda:out" />
</route>

Sample for update:

Java DSL:

from("direct:update")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.UPDATE_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:update" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>update</constant>
    </setHeader>
    <to uri="hazelcast:map:foo" />
</route>

Sample for delete:

Java DSL:

from("direct:delete")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.DELETE_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:delete" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>delete</constant>
    </setHeader>
    <to uri="hazelcast:map:foo" />
</route>

Sample for query

Java DSL:

from("direct:query")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.QUERY_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX)
.to("seda:out");

Spring DSL:

<route>
    <from uri="direct:query" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>query</constant>
    </setHeader>
    <to uri="hazelcast:map:foo" />
    <to uri="seda:out" />
</route>

For the query operation Hazelcast offers a SQL like syntax to query your distributed map.

String q1 = "bar > 1000";
template.sendBodyAndHeader("direct:query", null, HazelcastConstants.QUERY, q1);

Map cache consumer - from("hazelcast:map:foo")

Hazelcast provides event listeners on their data grid. If you want to be notified if a cache will be manipulated, you can use the map consumer. There’re 4 events: put, update, delete and envict. The event type will be stored in the "hazelcast.listener.action" header variable. The map consumer provides some additional information inside these variables:

Header Variables inside the response message:

Name Type Description

CamelHazelcastListenerTime

Long

time of the event in millis

CamelHazelcastListenerType

String

the map consumer sets here "cachelistener"

CamelHazelcastListenerAction

String

type of event - here added, updated, envicted and removed.

CamelHazelcastObjectId

String

the oid of the object

CamelHazelcastCacheName

String

the name of the cache - e.g. "foo"

CamelHazelcastCacheType

String

the type of the cache - here map

The object value will be stored within put and update actions inside the message body.

Here’s a sample:

fromF("hazelcast:%sfoo", HazelcastConstants.MAP_PREFIX)
.log("object...")
.choice()
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ADDED))
         .log("...added")
         .to("mock:added")
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ENVICTED))
         .log("...envicted")
         .to("mock:envicted")
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.UPDATED))
         .log("...updated")
         .to("mock:updated")
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.REMOVED))
         .log("...removed")
         .to("mock:removed")
    .otherwise()
         .log("fail!");

Usage of Multi Map

multimap cache producer - to("hazelcast:multimap:foo")

A multimap is a cache where you can store n values to one key. The multimap producer provides 4 operations (put, get, removevalue, delete).

Header Variables for the request message:

Name Type Description

CamelHazelcastOperationType

String

valid values are: put, get, removevalue, delete From Camel 2.16: clear.

CamelHazelcastObjectId

String

the object id to store / find your object inside the cache

Sample for put:

Java DSL:

from("direct:put")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PUT_OPERATION))
.to(String.format("hazelcast:%sbar", HazelcastConstants.MULTIMAP_PREFIX));

Spring DSL:

<route>
    <from uri="direct:put" />
    <log message="put.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>put</constant>
    </setHeader>
    <to uri="hazelcast:multimap:foo" />
</route>

Sample for removevalue:

Java DSL:

from("direct:removevalue")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.REMOVEVALUE_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.MULTIMAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:removevalue" />
    <log message="removevalue..."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>removevalue</constant>
    </setHeader>
    <to uri="hazelcast:multimap:foo" />
</route>

To remove a value you have to provide the value you want to remove inside the message body. If you have a multimap object \{key: "4711" values: { "my-foo", "my-bar"}} you have to put "my-foo" inside the message body to remove the "my-foo" value.

Sample for get:

Java DSL:

from("direct:get")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.GET_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.MULTIMAP_PREFIX)
.to("seda:out");

Spring DSL:

<route>
    <from uri="direct:get" />
    <log message="get.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>get</constant>
    </setHeader>
    <to uri="hazelcast:multimap:foo" />
    <to uri="seda:out" />
</route>

Sample for delete:

Java DSL:

from("direct:delete")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.DELETE_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.MULTIMAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:delete" />
    <log message="delete.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>delete</constant>
    </setHeader>
    <to uri="hazelcast:multimap:foo" />
</route>

you can call them in your test class with:

template.sendBodyAndHeader("direct:[put|get|removevalue|delete]", "my-foo", HazelcastConstants.OBJECT_ID, "4711");

multimap cache consumer - from("hazelcast:multimap:foo")

For the multimap cache this component provides the same listeners / variables as for the map cache consumer (except the update and enviction listener). The only difference is the multimap prefix inside the URI. Here is a sample:

fromF("hazelcast:%sbar", HazelcastConstants.MULTIMAP_PREFIX)
.log("object...")
.choice()
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ADDED))
        .log("...added")
                .to("mock:added")
        //.when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ENVICTED))
        //        .log("...envicted")
        //        .to("mock:envicted")
        .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.REMOVED))
                .log("...removed")
                .to("mock:removed")
        .otherwise()
                .log("fail!");

Header Variables inside the response message:

Name Type Description

CamelHazelcastListenerTime

Long

time of the event in millis

CamelHazelcastListenerType

String

the map consumer sets here "cachelistener"

CamelHazelcastListenerAction

String

type of event - here added and removed (and soon envicted)

CamelHazelcastObjectId

String

the oid of the object

CamelHazelcastCacheName

String

the name of the cache - e.g. "foo"

CamelHazelcastCacheType

String

the type of the cache - here multimap

Usage of Queue

Queue producer – to(“hazelcast:queue:foo”)

The queue producer provides 6 operations (add, put, poll, peek, offer, removevalue).

Sample for add:

from("direct:add")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.ADD_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.QUEUE_PREFIX);

Sample for put:

from("direct:put")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PUT_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.QUEUE_PREFIX);

Sample for poll:

from("direct:poll")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.POLL_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.QUEUE_PREFIX);

Sample for peek:

from("direct:peek")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PEEK_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.QUEUE_PREFIX);

Sample for offer:

from("direct:offer")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.OFFER_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.QUEUE_PREFIX);

Sample for removevalue:

from("direct:removevalue")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.REMOVEVALUE_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.QUEUE_PREFIX);

Queue consumer – from(“hazelcast:queue:foo”)

The queue consumer provides 2 operations (add, remove).

fromF("hazelcast:%smm", HazelcastConstants.QUEUE_PREFIX)
   .log("object...")
   .choice()
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ADDED))
            .log("...added")
        .to("mock:added")
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.REMOVED))
        .log("...removed")
        .to("mock:removed")
    .otherwise()
        .log("fail!");

Usage of Topic

Topic producer – to(“hazelcast:topic:foo”)

The topic producer provides only one operation (publish).

Sample for publish:

from("direct:add")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PUBLISH_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.PUBLISH_OPERATION);

Topic consumer – from(“hazelcast:topic:foo”)

The topic consumer provides only one operation (received). This component is supposed to support multiple consumption as it’s expected when it comes to topics so you are free to have as much consumers as you need on the same hazelcast topic.

fromF("hazelcast:%sfoo", HazelcastConstants.TOPIC_PREFIX)
  .choice()
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.RECEIVED))
      .log("...message received")
    .otherwise()
      .log("...this should never have happened")
 

Usage of List

List producer – to(“hazelcast:list:foo”)

The list producer provides 4 operations (add, addAll, set, get, removevalue, removeAll, clear).

Sample for add:

from("direct:add")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.ADD_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.LIST_PREFIX);

Sample for get:

from("direct:get")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.GET_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.LIST_PREFIX)
.to("seda:out");

Sample for setvalue:

from("direct:set")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.SETVALUE_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.LIST_PREFIX);

Sample for removevalue:

from("direct:removevalue")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.REMOVEVALUE_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.LIST_PREFIX);

Note that CamelHazelcastObjectIndex header is used for indexing purpose.

The list consumer provides 2 operations (add, remove).List consumer – from(“hazelcast:list:foo”)

fromF("hazelcast:%smm", HazelcastConstants.LIST_PREFIX)
    .log("object...")
    .choice()
        .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ADDED))
            .log("...added")
                        .to("mock:added")
        .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.REMOVED))
            .log("...removed")
                        .to("mock:removed")
                .otherwise()
                        .log("fail!");

Usage of SEDA

SEDA component differs from the rest components provided. It implements a work-queue in order to support asynchronous SEDA architectures, similar to the core "SEDA" component.

SEDA producer – to(“hazelcast:seda:foo”)

The SEDA producer provides no operations. You only send data to the specified queue.

Name Default value Description

transferExchange

false

if set to true the whole Exchange will be transfered. If header or body contains not serializable objects, they will be skipped.

Java DSL :

from("direct:foo")
.to("hazelcast:seda:foo");

Spring DSL :

<route>
   <from uri="direct:start" />
   <to uri="hazelcast:seda:foo" />
</route>

SEDA consumer – from(“hazelcast:seda:foo”)

The SEDA consumer provides no operations. You only retrieve data from the specified queue.

Name Default value Description

pollInterval

1000

The timeout used when consuming from the SEDA queue. When a timeout occurs, the consumer can check whether it is allowed to continue running. Setting a lower value allows the consumer to react more quickly upon shutdown. (deprecated from Camel 2.15 onwards, use pollTimeout instead).

pollTimeout

1000

The timeout used when consuming from the SEDA queue. When a timeout occurs, the consumer can check whether it is allowed to continue running. Setting a lower value allows the consumer to react more quickly upon shutdown.

concurrentConsumers

1

To use concurrent consumers polling from the SEDA queue.

transferExchange

false

if set to true the whole Exchange will be transfered. If header or body contains not serializable objects, they will be skipped.

transacted

false

if set to true then the consumer runs in transaction mode, where the messages in the seda queue will only be removed if the transaction commits, which happens when the processing is complete.

Java DSL :

from("hazelcast:seda:foo")
.to("mock:result");

Spring DSL:

<route>
  <from uri="hazelcast:seda:foo" />
  <to uri="mock:result" />
</route>

Usage of Atomic Number

  • There is no consumer for this endpoint! *

atomic number producer - to("hazelcast:atomicnumber:foo")

An atomic number is an object that simply provides a grid wide number (long). The operations for this producer are setvalue (set the number with a given value), get, increase (+1), decrease (-1) and destroy.

Header Variables for the request message:

Name Type Description

CamelHazelcastOperationType

String

valid values are: setvalue, get, increase, decrease, destroy

Sample for set:

Java DSL:

from("direct:set")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.SETVALUE_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.ATOMICNUMBER_PREFIX);

Spring DSL:

<route>
    <from uri="direct:set" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>setvalue</constant>
    </setHeader>
    <to uri="hazelcast:atomicvalue:foo" />
</route>

Provide the value to set inside the message body (here the value is 10): template.sendBody("direct:set", 10);

Sample for get:

Java DSL:

from("direct:get")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.GET_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.ATOMICNUMBER_PREFIX);

Spring DSL:

<route>
    <from uri="direct:get" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>get</constant>
    </setHeader>
    <to uri="hazelcast:atomicvalue:foo" />
</route>

You can get the number with long body = template.requestBody("direct:get", null, Long.class);.

Sample for increment:

Java DSL:

from("direct:increment")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.INCREMENT_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.ATOMICNUMBER_PREFIX);

Spring DSL:

<route>
    <from uri="direct:increment" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>increment</constant>
    </setHeader>
    <to uri="hazelcast:atomicvalue:foo" />
</route>

The actual value (after increment) will be provided inside the message body.

Sample for decrement:

Java DSL:

from("direct:decrement")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.DECREMENT_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.ATOMICNUMBER_PREFIX);

Spring DSL:

<route>
    <from uri="direct:decrement" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>decrement</constant>
    </setHeader>
    <to uri="hazelcast:atomicvalue:foo" />
</route>

The actual value (after decrement) will be provided inside the message body.

Sample for destroy

Java DSL:

from("direct:destroy")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.DESTROY_OPERATION))
.toF("hazelcast:%sfoo", HazelcastConstants.ATOMICNUMBER_PREFIX);

Spring DSL:

<route>
    <from uri="direct:destroy" />
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>destroy</constant>
    </setHeader>
    <to uri="hazelcast:atomicvalue:foo" />
</route>

cluster support

  • This endpoint provides no producer! *

instance consumer - from("hazelcast:instance:foo")

Hazelcast makes sense in one single "server node", but it’s extremly powerful in a clustered environment. The instance consumer fires if a new cache instance will join or leave the cluster.

Here’s a sample:

fromF("hazelcast:%sfoo", HazelcastConstants.INSTANCE_PREFIX)
.log("instance...")
.choice()
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ADDED))
        .log("...added")
        .to("mock:added")
    .otherwise()
        .log("...removed")
        .to("mock:removed");

Each event provides the following information inside the message header:

Header Variables inside the response message:

Name Type Description

CamelHazelcastListenerTime

Long

time of the event in millis

CamelHazelcastListenerType

String

the map consumer sets here "instancelistener"

CamelHazelcastListenerAction

String

type of event - here added or removed.

CamelHazelcastInstanceHost

String

host name of the instance

CamelHazelcastInstancePort

Integer

port number of the instance

Using hazelcast reference

By its name

<bean id="hazelcastLifecycle" class="com.hazelcast.core.LifecycleService"
      factory-bean="hazelcastInstance" factory-method="getLifecycleService"
      destroy-method="shutdown" />

<bean id="config" class="com.hazelcast.config.Config">
    <constructor-arg type="java.lang.String" value="HZ.INSTANCE" />
</bean>

<bean id="hazelcastInstance" class="com.hazelcast.core.Hazelcast" factory-method="newHazelcastInstance">
    <constructor-arg type="com.hazelcast.config.Config" ref="config"/>
</bean>
<camelContext xmlns="http://camel.apache.org/schema/spring">
    <route id="testHazelcastInstanceBeanRefPut">
        <from uri="direct:testHazelcastInstanceBeanRefPut"/>
        <setHeader headerName="CamelHazelcastOperationType">
            <constant>put</constant>
        </setHeader>
        <to uri="hazelcast:map:testmap?hazelcastInstanceName=HZ.INSTANCE"/>
    </route>

    <route id="testHazelcastInstanceBeanRefGet">
        <from uri="direct:testHazelcastInstanceBeanRefGet" />
        <setHeader headerName="CamelHazelcastOperationType">
            <constant>get</constant>
        </setHeader>
        <to uri="hazelcast:map:testmap?hazelcastInstanceName=HZ.INSTANCE"/>
        <to uri="seda:out" />
    </route>
</camelContext>

By instance

<bean id="hazelcastInstance" class="com.hazelcast.core.Hazelcast"
      factory-method="newHazelcastInstance" />
<bean id="hazelcastLifecycle" class="com.hazelcast.core.LifecycleService"
      factory-bean="hazelcastInstance" factory-method="getLifecycleService"
      destroy-method="shutdown" />

<camelContext xmlns="http://camel.apache.org/schema/spring">
    <route id="testHazelcastInstanceBeanRefPut">
        <from uri="direct:testHazelcastInstanceBeanRefPut"/>
        <setHeader headerName="CamelHazelcastOperationType">
            <constant>put</constant>
        </setHeader>
        <to uri="hazelcast:map:testmap?hazelcastInstance=#hazelcastInstance"/>
    </route>

    <route id="testHazelcastInstanceBeanRefGet">
        <from uri="direct:testHazelcastInstanceBeanRefGet" />
        <setHeader headerName="CamelHazelcastOperationType">
            <constant>get</constant>
        </setHeader>
        <to uri="hazelcast:map:testmap?hazelcastInstance=#hazelcastInstance"/>
        <to uri="seda:out" />
    </route>
</camelContext>

Publishing hazelcast instance as an OSGI service

If operating in an OSGI container and you would want to use one instance of hazelcast across all bundles in the same container. You can publish the instance as an OSGI service and bundles using the cache al need is to reference the service in the hazelcast endpoint.

Bundle A create an instance and publishes it as an OSGI service

 
<bean id="config" class="com.hazelcast.config.FileSystemXmlConfig">
    <argument type="java.lang.String" value="${hazelcast.config}"/>
</bean>

<bean id="hazelcastInstance" class="com.hazelcast.core.Hazelcast" factory-method="newHazelcastInstance">
    <argument type="com.hazelcast.config.Config" ref="config"/>
</bean>

<!-- publishing the hazelcastInstance as a service -->
<service ref="hazelcastInstance" interface="com.hazelcast.core.HazelcastInstance" />

Bundle B uses the instance

<!-- referencing the hazelcastInstance as a service -->
<reference ref="hazelcastInstance" interface="com.hazelcast.core.HazelcastInstance" />

<camelContext xmlns="http://camel.apache.org/schema/blueprint">
    <route id="testHazelcastInstanceBeanRefPut">
        <from uri="direct:testHazelcastInstanceBeanRefPut"/>
        <setHeader headerName="CamelHazelcastOperationType">
            <constant>put</constant>
        </setHeader>
        <to uri="hazelcast:map:testmap?hazelcastInstance=#hazelcastInstance"/>
    </route>

    <route id="testHazelcastInstanceBeanRefGet">
        <from uri="direct:testHazelcastInstanceBeanRefGet" />
        <setHeader headerName="CamelHazelcastOperationType">
            <constant>get</constant>
        </setHeader>
        <to uri="hazelcast:map:testmap?hazelcastInstance=#hazelcastInstance"/>
        <to uri="seda:out" />
    </route>
</camelContext>

Usage of Replicated map

Avalaible from Camel 2.16

replicatedmap cache producer

A replicated map is a weakly consistent, distributed key-value data structure with no data partition. The replicatedmap producer provides 4 operations (put, get, delete, clear).

Header Variables for the request message:

Name Type Description

CamelHazelcastOperationType

String

valid values are: put, get, removevalue, delete

CamelHazelcastObjectId

String

the object id to store / find your object inside the cache

Sample for put:

Java DSL:

from("direct:put")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.PUT_OPERATION))
.to(String.format("hazelcast:%sbar", HazelcastConstants.REPLICATEDMAP_PREFIX));

Spring DSL:

<route>
    <from uri="direct:put" />
    <log message="put.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>put</constant>
    </setHeader>
    <to uri="hazelcast:replicatedmap:foo" />
</route>

Sample for get:

Java DSL:

from("direct:get")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.GET_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.REPLICATEDMAP_PREFIX)
.to("seda:out");

Spring DSL:

<route>
    <from uri="direct:get" />
    <log message="get.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>get</constant>
    </setHeader>
    <to uri="hazelcast:replicatedmap:foo" />
    <to uri="seda:out" />
</route>

Sample for delete:

Java DSL:

from("direct:delete")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.DELETE_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.REPLICATEDMAP_PREFIX);

Spring DSL:

<route>
    <from uri="direct:delete" />
    <log message="delete.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>delete</constant>
    </setHeader>
    <to uri="hazelcast:replicatedmap:foo" />
</route>

you can call them in your test class with:

template.sendBodyAndHeader("direct:[put|get|delete|clear]", "my-foo", HazelcastConstants.OBJECT_ID, "4711");

replicatedmap cache consumer

For the multimap cache this component provides the same listeners / variables as for the map cache consumer (except the update and enviction listener). The only difference is the multimap prefix inside the URI. Here is a sample:

fromF("hazelcast:%sbar", HazelcastConstants.MULTIMAP_PREFIX)
.log("object...")
.choice()
    .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ADDED))
        .log("...added")
                .to("mock:added")
        //.when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.ENVICTED))
        //        .log("...envicted")
        //        .to("mock:envicted")
        .when(header(HazelcastConstants.LISTENER_ACTION).isEqualTo(HazelcastConstants.REMOVED))
                .log("...removed")
                .to("mock:removed")
        .otherwise()
                .log("fail!");

Header Variables inside the response message:

Name Type Description

CamelHazelcastListenerTime

Long

time of the event in millis

CamelHazelcastListenerType

String

the map consumer sets here "cachelistener"

CamelHazelcastListenerAction

String

type of event - here added and removed (and soon envicted)

CamelHazelcastObjectId

String

the oid of the object

CamelHazelcastCacheName

String

the name of the cache - e.g. "foo"

CamelHazelcastCacheType

String

the type of the cache - here replicatedmap

Usage of Ringbuffer

Avalaible from Camel 2.16

ringbuffer cache producer

Ringbuffer is a distributed data structure where the data is stored in a ring-like structure. You can think of it as a circular array with a certain capacity. The ringbuffer producer provides 5 operations (add, readonceHead, readonceTail, remainingCapacity, capacity).

Header Variables for the request message:

Name Type Description

CamelHazelcastOperationType

String

valid values are: put, get, removevalue, delete

CamelHazelcastObjectId

String

the object id to store / find your object inside the cache

Sample for put:

Java DSL:

from("direct:put")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.ADD_OPERATION))
.to(String.format("hazelcast:%sbar", HazelcastConstants.RINGBUFFER_PREFIX));

Spring DSL:

<route>
    <from uri="direct:put" />
    <log message="put.."/>
        <!-- If using version 2.8 and above set headerName to "CamelHazelcastOperationType" -->
    <setHeader headerName="hazelcast.operation.type">
        <constant>add</constant>
    </setHeader>
    <to uri="hazelcast:ringbuffer:foo" />
</route>

Sample for readonce from head:

Java DSL:

from("direct:get")
.setHeader(HazelcastConstants.OPERATION, constant(HazelcastConstants.READ_ONCE_HEAD_OPERATION))
.toF("hazelcast:%sbar", HazelcastConstants.RINGBUFFER_PREFIX)
.to("seda:out");