I'm relatively new to spring integration but I have been tasked with implementing a tcp gateway that needs to:
Listen for messages on a socket
Receive a message
Process the message and write some data to a queue
Return a response message to the original socket.
In my experience with spring integration, the message flow has not been bi-directional. I've only configured routers to listen, handle messages, and output to queue/topic. In this case, though, I need to accept messages and return a response while also forwarding on some message to a queue. Suggestions?
This is my integration xml so far.
<int:chain input-channel="tcpChannel">
<int:service-activator ref="tcpHandler" method="handleInput" />
</int:chain>
<int-ip:tcp-inbound-gateway
id="tcpGateway"
connection-factory="tcpServer"
request-channel="tcpChannel" />
How can I have the output of handleInput be forwarded to a queue but also have some response sent back from the gateway?
Edit: After the conversation below with Gary, this seems to be the pattern we want to follow:
<int-ip:tcp-inbound-gateway id="tcpGateway"
connection-factory="tcpServer"
request-channel="tcpChannel"
reply-channel="tcpReplyChannel"/>
<int:publish-subscribe-channel id="tcpChannel" />
<int:chain input-channel="tcpChannel">
<!-- int:json-to-object-transformer type="com.heb.revo.events.RxPosCredit" /-->
<int:service-activator ref="tcpHandler" method="handleInputToQueue" />
<jms:outbound-channel-adapter destination-name="${queue.response}" />
</int:chain>
<int:service-activator id="tcpResponseHandler"
ref="tcpHandler" method="replyToSocket"
input-channel="tcpChannel"
output-channel="tcpReplyChannel" />
<int:publish-subscribe-channel id="tcpReplyChannel" />
Since your chain has no output-channel the framework will route the return value of the handleInput method back to the gateway automatically.
If you want to capture the result and send it somewhere else (as well as a reply), create a <int:publish-subscribe-channel id="foo"/>, set the output-channel of the chain to foo, set the reply-channel on the gateway to foo, and subscribe another endpoint to foo (as an input-channel.
Related
There is one request queue and the reply queues are created by the client server instances and pinned to each instance rather than using temporary queues.
The use case needs to get an inbound jms message and then send that message to an asynchronous process. Once the async reply messsage is received from the service I need to take those results and reply back to the original message's jmsReplyTo. The Jms gateway would not work in this instance AFAIK>
I am using a jms message driven channel adapter for the message in with a series of channels and service activators to handle the out of process calls and async replies. I am trying to use the DynamicDestinationResolver to no avail. Additionally I have tried to set the outbound destination address programatically but could not figure out a good way to do this.
This seems like a common pattern but I could not find a good example for a completely disconnected async request response. Disconnected meaning that the usual async jms request reply did not seem to fit the need.
Context Config:
<!-- Input from Amq -->
<amq:queue id="requestQueue" physicalName="${request.queue}" />
<int-jms:message-driven-channel-adapter id="jmsIn"
connection-factory="jmsConnectionFactory"
destination="requestQueue"
channel="queueRequestChannel" concurrent-consumers="5" />
<int:channel id="queueRequestChannel" />
<int:service-activator input-channel="queueRequestChannel" ref="switchMessageHandler" method="processSwitchMessage"
output-channel="cardNetworkOutChannel"/>
<!-- Output to Card Network-->
<int:channel id="cardNetworkOutChannel" />
<!--<int:service-activator input-channel="cardNetworkOutChannel" ref="cardNetworkHandler" method="send8583Message" />-->
<!-- Simply used to mock the card network by transforming a SwithMessage to a SwitchMessageResponse * Not needed for target solution -->
<int:transformer id="requestResponseTransformer" ref="nettyCardNetworkClientMock" input-channel="cardNetworkOutChannel"
method="process" output-channel="cardNetworkInChannel"/>
<!-- Input from Card Network -->
<int:channel id="cardNetworkInChannel" />
<int:service-activator input-channel="cardNetworkInChannel" ref="switchMessageHandler" method="sendSwitchMessage"
output-channel="queueReplyChannel"/>
<int:channel id="queueReplyChannel"/>
<int-jms:outbound-channel-adapter
destination-resolver="simpleDestinationResolver" connection-factory="jmsConnectionFactory"
channel="queueReplyChannel" destination-expression="headers.jms_replyTo" />
I just updated the jms sample app to make the server side use independent adapters instead of the inbound gateway and it works just fine...
<!-- <jms:inbound-gateway id="jmsin" -->
<!-- request-destination="requestQueue" -->
<!-- request-channel="demoChannel"/> -->
<channel id="demoChannel"/>
<jms:message-driven-channel-adapter destination="requestQueue" channel="demoChannel" />
<service-activator input-channel="demoChannel" ref="demoBean" output-channel="reply" />
<channel id="reply" />
<jms:outbound-channel-adapter channel="reply" destination-expression="headers['jms_replyTo']" />
Turn on DEBUG logging - we put out lots of useful stuff.
EDIT
I just made it async by...
<channel id="reply">
<queue/>
</channel>
<jms:outbound-channel-adapter channel="reply" destination-expression="headers['jms_replyTo']">
<poller fixed-delay="3000"/>
</jms:outbound-channel-adapter>
EDIT2
Depending on what you are using on the client side, many clients require the inbound message id to be used as the correlation id. (This is true for the outbound gateway by default, with named reply queue, unless you provide a correlation-key).
So, to set up the correlationId, you can use a header enricher; I just tested this...
<chain input-channel="reply">
<header-enricher>
<header name="jms_correlationId" expression="headers['jms_messageId']" />
</header-enricher>
<jms:outbound-channel-adapter destination-expression="headers['jms_replyTo']"/>
<poller fixed-delay="1000" />
</chain>
This is not the issue if the client side is setting the correlation id header itself.
I have a jms-inbound-gateway that reads requests from a WebsphereMQ broker, passes them though my integration system and then replies with a response message.
I need to log the messages with the jms_messageId and jms_correlationId headers set, so I can match request/reply messages in the log file (and show it to my client when he says that my response does not have the correct jms_correlationId)
Is there a way to intercept the method producer.sendReply(...) after the correlationId header is set?
There is no need to intercept there; the headers are available in the Spring Integration messsage in the gateway reply message before it gets to the gateway.
Simply make the reply-channel a publish-subscribe-channel and add a <logging-channel-adapter/> that has it as its input channel.
The reply message will be sent to both the gateway and the logger.
If you are using the default mechanism to route the reply (no output-channel on your last integration component), simply add an output-channel and route to the reply channel.
This is the gist of my solution after Gary's input:
<jms:inbound-gateway
id="inboundDestination"
connection-factory="connectionFactory"
request-destination="nmRequestsQueue"
request-channel="request-begin"
reply-channel="request-end"
error-channel="normaErrorChannel"
concurrent-consumers="1"
acknowledge="transacted" />
<int:logging-channel-adapter id="request-response-logger"
log-full-message="true"
level="DEBUG"
logger-name="com.audaxys.si.messages" />
<int:channel id="request-begin">
<int:interceptors>
<int:wire-tap channel="request-response-logger" />
</int:interceptors>
</int:channel>
<int:chain input-channel="request-begin" output-channel="request-end">
... Do Stuff ...
</int:chain>
<int:publish-subscribe-channel id="request-end">
<int:interceptors>
<int:wire-tap channel="request-response-logger" />
</int:interceptors>
</int:publish-subscribe-channel>
I have http outbound gateway that post json message to rest service, now the rest will response with http error status with json message type in case if there is any error which should captured by our application. in happy scenario which post json message to that rest service and got http success status and json message also should captured with our application.
Now what I achieved through spring integration I was able to send message and got success response and capture it but in error status the spring behavior it throw an exception how can I change the behavior?
also how can make retry if there is any timeout ?
below the http outbound gateway configuration
<int-http:outbound-gateway request-channel="aggregatorChannel" reply-channel="responseChannel" charset="UTF-8"
url="http://localhost:8090/receiveGateway" http-method="POST" reply-timeout="180000">
<int-http:request-handler-advice-chain >
<int:retry-advice max-attempts="5" recovery-channel="aggregatorChannel" >
<int:exponential-back-off initial="1000" multiplier="5.0" maximum="600000" />
</int:retry-advice>
</int-http:request-handler-advice-chain>
</int-http:outbound-gateway>
the above configuration will retry on the error message which I don't want retry on that, I want to retry if there is only timeout.
What do you want to change the behavior to (when there's an error)?
If you wire up the RetryTemplate for the retry interceptor manually, as a <bean/>, you can specify in the RetryPolicy which exceptions are retryable.
EDIT:
It's not entirely clear what you are trying to do, but maybe this will point you in the right direction...
<int-http:outbound-gateway request-channel="requestChannel"
url="http://localhost:8080/http/receiveGateway"
http-method="POST"
request-factory="requestFactory"
expected-response-type="java.lang.String">
<int-http:request-handler-advice-chain >
<int:retry-advice max-attempts="5" recovery-channel="foo">
<int:exponential-back-off initial="1000" multiplier="1.1" maximum="600000" />
</int:retry-advice>
<bean class="foo.MyExceptionAdvice" />
</int-http:request-handler-advice-chain>
</int-http:outbound-gateway>
<bean id="requestFactory" class="org.springframework.http.client.SimpleClientHttpRequestFactory">
<property name="connectTimeout" value="5000"/>
<property name="readTimeout" value="5000"/>
</bean>
If the custom exception advice throws an exception, the message will be retried and eventually sent to he recovery channel. If the custom advice consumes the error, it can send a return value. I have put a simple on this gist.
I want to create programmatically the following XML config on demand:
<int-mqtt:message-driven-channel-adapter id="inboundAdapter"
client-id="${mqtt.client.id}"
url="${mqtt.broker.url}"
topics="${mqtt.subscribed.topics}"
client-factory="clientFactory"
channel="input-channel-1" converter="customConverter" />
<int:channel id="input-channel-1">
<int:queue/>
<int:interceptors>
<int:wire-tap channel="logger"/>
<int:ref bean="messageListener"/>
</int:interceptors>
</int:channel>
<int:channel id="logger" />
<int:logging-channel-adapter channel="logger"
auto-startup="true" level="INFO" id="loggerAdapter" log-full-message="true" />
What i can do is the following
CustomMqttPahoMessageDrivenChannelAdapter adapter = new CustomMqttPahoMessageDrivenChannelAdapter(url, clientId, topic);
adapter.setOutputChannel(outputChannel);
adapter.setConverter(ctx.getBean("customConverter", MyPahoMessageConverter.class));
Now I need to add interceptor bean through which each client will get notified when message arrives according to their subscribed topics respectively.
What I am trying to achieve is:
1) Create mqtt adapter when a client connects to server.(each client will subscribe to different topics as per configuration)
2) Dispose mqtt adapter when client disconnects.
Can anyone help me on this?
It's not clear what you are trying to do; what do you have downstream of the input-channel-1 in the XML configuration.
What does messageListener do?
It is an anti-pattern to put business logic in a channel; unless it's something really lightweight, consider invoking it using a <service-activator/> instead - possibly by making input-channel-1 a pub-sub channel.
To answer your simple question, to add the interceptor, you can use outputChannel.addInterceptor(ctx.getBean("messageListener", ChannelInterceptor.class));.
we are using Spring Integration 2.1 for persisting messages into database sent by clients.
There is a queue which will be filled by a custom adapter. The configured service activator polls this queue and releases the message to a spring managed #Repository bean. All errors will be captured to an error channel and will be handled by a service. The configuration works so far fine.
My concern is that if the database is not available the service-activators polls all incoming message from the queue and puts them into the error channel. Is there a way to prevent the service-activator to poll the message if the database is obviously not available, for example by sending a test query ?
My configuraton:
<int:channel id="inChannel">
<int:queue />
</int:channel>
<bean id="service" class="some.service.Service" />
<int:service-activator ref="service"
method="write" input-channel="inChannel">
<int:poller fixed-rate="100" task-executor="srvTaskExecutor"
receive-timeout="90" error-channel="errChannel" />
</int:service-activator>
<task:executor id="srvTaskExecutor" pool-size="2-10"
queue-capacity="0" rejection-policy="DISCARD" />
<int:channel id="errChannel" />
<int:service-activator input-channel="errChannel"
ref="errorService" method="write"/>
Regards.
If you give the polling service-activator an "id", you can refer to that instance and call start() or stop() on it based on the DB being available or not. Most likely you'd want to set auto-startup="false" on that service-activator as well.
Additionally, you can even define a "control-bus" element and then send messages like "myActivator.start()" and "myActivator.stop()" to that control bus' input-channel.
Hope that helps,
Mark