I'm working on my first Camel-based integration project and have a question about integrating to an existing TCP-server (for which I don't have a source code so no changes can be done to it) with either Mina or Netty (not sure which is the better one to use for this). I've read the Camel in Action-book but didn't find the answer there for this particular scenario.
The scenario is as follows: - There is a legacy server application X that exposes a TCP server-port 9999. - The application X is to be used so that a TCP-client opens a connection to port 9999, sends one string "CONNECT"+CRLF and nothing else and then starts listening the incoming stream from the server. - Application X will periodically (time varies randomly, from seconds to hours) send a string to the TCP-client and the TCP-client should then do something with it, in my case it should be forwarded to a JMS-queue as a TextMessage. - The TCP-client should keep the connection open indefinitely and it connection is lost, automatically retry the connection until it succeeds and then continue processing like above. I have been able to partly implement this with Camel with Mina BUT ONLY so that Mina exposes a TCP-server that can be contacted with my simple test-client. I am able to send strings from the test client and have them forwarded as JMS-messages so the routing part seems to work as long as I just get something into the TCP-server endpoint. But here is my problem How do I get the TCP-endpoint to act as TCP-client instead of TCP-server, i.e. connect to the TCP-server port 9999, send the CONNECT-string once and then just keep the connection open and process the incoming strings from the TCP-server?!? I can't use polling to contact the server because it doesn't store anything for later retrieval, it will only send data to connected TCP-clients if they are connect at the moment when it has data to send, it is a kind of "sensor"-type of application with no message/data buffering. Furthermore, the system is very time-sensitive, I must process the text message immediately after the server sends it, say, within 100-200ms, definitely under 1s. So, first of all: Is this possible with existing components, either Mina or Netty? Which one? How would I define the endpoint? Or do I need to write my own component? I've done some TCP-socket programming with Java so that would be doable, but I'd like to avoid that if possible and just use the existing stuff provided with Camel, otherwise; what's the point of using it. But obviously I'll try the custom-component way if I have to. So optimally I'd like the system to act something like this: - When I start the CamelContext, a TCP-client endpoint will actively try to connect to the TCP-server on port 9999 and if it fails, it will try again periodically, say, once every 10 seconds. And also if the connection is lost, it would retry to connect again automatically. The reason being that the server may be unreachable at times so the TCP-client endpoint should be robust enough to retry to connect if the connection is lost or server app not reached immediately. - After connection is formed, the TCP-client endpoint should keep the connection open, listen to the inbound server-stream and when it receives a line of text from the server, just forward it to the JMS-queue as textmessage,the received string being the body. Any advice on implementing this? It doesn't feel to me that complex scenario, only the TCP-client/server roles seem a bit problematic. -- View this message in context: http://camel.465427.n5.nabble.com/How-to-act-as-TCP-client-with-Camel-Mina-Netty-tp5718671.html Sent from the Camel - Users mailing list archive at Nabble.com.
