Send messages to the temporary queue with MassTransit - masstransit

I have the following scenario, where i can not respond directly to the request of a MassTransit RequestClient
Request Service 1 implementing a RequestClient and calling GetResponse
Another service 2 have the consumer of the request that just save the request information (messageId, requestId ...) to some storage and not send a response back.
Based on an API call in service 2, we can find the saved request information and send back the response to the RequestClient. I could achive this by sending the response to the temporaty queue of the RequestClient bus but the message ends up in the skipped queue of the temporary queue and GetResponse never gets the message.
How can i make the message arrives to the main temporary queue, and that the GetResponse consumer find it.

If anyone has a similar scenario. We achived this by implementing an event waiter mechanism in the consumer. When the external API call is made in service 2, an internal event is fired then consumer who is waiting for it through ResetEvent continue by gathering the response and send it back through out of the box RespondAsync.

Related

setting ack to client results in duplicate message delivery stomp

I have a stomp application and I set the ack mode to client during subscription from front-end. During the subscription the same client during a session would connect to the same queue in the message broker ActiveMQ Artemis because a session id is used for queue.But a different id is passed in the header every time the user reloads the page.
The subscription code is shown below.
consumer.subscribe(
'/queue.'+ rscSessionId,
function (response) {
console.log("resposne");
},
{
"subscription-type": "ANYCAST",
ack: "client",
id: generateId(),
}
);
Once I send message from client it gets send to frontend with following headers.
MESSAGE
subscription:1234
message-id:1543449
destination:/queue.sessionid
expires:1653914606725
redelivered:false
priority:4
persistent:true
timestamp:1653912806717
destination-type:ANYCAST
__AMQ_CID:30e64793-dd9b-11ec-8843-c238460c3152
_type:ResponseData
content-length:300
and afterwards on every reload.
MESSAGE
subscription:1235
message-id:1543449
destination:/queue.sessionid
expires:1653914606725
redelivered:true
priority:4
persistent:true
timestamp:1653912806717
destination-type:ANYCAST
__AMQ_CID:30e64793-dd9b-11ec-8843-c238460c3152
_type:ResponseData
content-length:300
You can see that redelivered is set to true after first delivery and after that every response has redelivered to false. the subscription value keeps getting changed though because of the id passed in header.
How can I make it deliver only once but with guarantee?
Are you actually acknowledging the message once you receive it? If you don't acknowledge the message then once the client closes the connection and subscribes again it will receive the same message again. The different value you're seeing for the subscription header indicates this is what is happening.
To be clear, if you're using the client acknowledgement mode then once you receive the MESSAGE frame you need to send a corresponding ACK frame in order to acknowledge the message. Your client should have a way to do this since it is a standard part of the STOMP specification.
I guess that Artemis is configured with publisher/subscriber messaging style. In this case, every new consumer/receiver/client will get the messages aka you write/read from a topic.
Changing to a point-to-point message style would fix the issue (would only be read once) aka a queue.
This is usually decided by the setup but I recall that ActiveMQ allowed producer to create those on the fly.
If you do need a topic, then you will have to handle the IDs on the client side.

In a group chat, should the new message event (websocket) be sent by the client or the API?

I have a doubt, in a group chat system that has a database with a rest API, who should issue the event of a new message?
The client or the endpoint to create the new message?
For example: X user sends a message to group Y, then uses the api endpoint api.com/message-create and that endpoint emits the message-create event through websocket
Example 2: X user sends a message to group Y, then uses the api api.com/message-create endpoint and the endpoint does not emit the message-create event, but was emitted when the user pressed the send message button
I still don't quite understand if it would occupy more websocket channels to achieve that, if a global one is enough, etc.
The server should be responsible for communication logic. So your first example is better.
But: why do you use two communication channels for sending an creating messages?
If you use websocket, you don't need create a message from a client by using additional rest endpoint.
This approach is prone to errors. For example if the client will loose network connection after sending message through websocket and before executing call to the REST endpoint?
The message will not be stored in a database.
Your flow should looks as follows:
User clicks the send button.
Message is send through the websocket.
Message is stored in the database asynchronously (you can do it directly from communication server, or use rest endpoint)
Emit "new message" event to the group.

Is it guaranteed that subscription for response is established before request is processed in temp-topic based request/response scheme?

I would like to implement an request/response pattern in the following parts:
Server: Springboot with ActiveMQ
Client: JavaScript with stompjs over websocket
I want a behaviour like an http request.
Send a Request, get a corresponding Response.
I try to do this with temporary channels
I send 3 Messages
The steps are:
SUBSCRIBE to a temprorary channel
SUBSCRIBE
id:mysub
destination:/temp-queue/example
SEND the request and include a reply-to header with the subscribed channel of Step 1
destination:/queue/PO.REQUEST
reply-to:/temp-queue/example
Get the Response Message from the Server
MESSAGE
subscription:foo
reply-to:/queue/temp.default.23.example
destination:/queue/PO.REQUEST
reply-to:/temp-queue/example
But now (As Client send messages asynchronous) im not sure if Step 1 is complete on server, and so server is ready to send Response to the queue when the Request of Step 2 arrives at the server.
Is it possible that server finishes Step 2 before finishing Step 1, and so sends the response to nowhere? Or does ActiveMQ ensures that the received messages 1 and 2 from the client are processed and finished in the correct order?
Can any race condition between message 1 and 2 happen?
Thank you very much!
Any STOMP frame that your client sends can be sent with a receipt request that makes the processing of that frame synchronous. So if you want to ensure that a subscribe is complete before doing the send, then attach a receipt-id to the subscribe frame and wait for the spec mandated RECEIPT frame before going on to do the send, this ensures that your subscription is setup prior to any other processing.

Spring Integration: Send response to client http inbound gateway

I have a http inbound gateway which needs to receive the request validate it and then immediately send response to client. After the response is sent back, my SI flow needs to continue with further processing. The response should be sent to client as soon as the validation is complete. Sending of response shouldn't wait until my entire processing is complete. How can I trigger SI flow to continue with further processing once the response is sent. What is the appropriate SI component for this scenario?
If the response is a simply 200 OK, use an inbound channel adapter (not a gateway) and make the first channel after the validation an ExecutorChannel. Then, immediately the message is handed off to the executor, the response will be sent.
If you need a custom reply, use a gateway and make the first channel after the validation a publish-subscribe channel (with an executor), construct the reply in one consumer of that channel, process the request in another.

What is the purpose of HandlerExtensions.ConnectHandle ConnectHandler<T>() method?

Xmldoc states:
Adds a message handler to the service bus for handling a specific type
of message
But it does not require endpoint name. How then does it work? I tried this method, but nothing happened.
Is there any possibility to add handlers dynamically, while bus is running?
By connecting a handler to the bus after it has been started, messages can be sent to the bus's address directly. This is particularly useful for things like responses to requests, which should not be published and are sent immediately back to the endpoint.
When using bus.ConnectHandler(context => {...}) to add a handler to the bus dynamically, no subscriptions or exchange bindings are created on the broker. It's only possible to receive messages which are directly sent to the endpoint.
When a message is sent from the bus, such as a request, the SourceAddress is added to the message header. If a request is sent, the ResponseAddress is also set. A fault address can also be specified if you want to use a non-dynamic endpoint to capture faults (such as a failed command that is not awaited, IE, fire and forget) so that faults can be triaged and handled appropriately by another persistent endpoint.

Resources