Continuing with my previous post on NServiceBus in this part I am going to implement what we had discuss about integration and how NServiceBus can be used. Based on the previous architecture diagram I am doing to drill into component architecture and for easy illustration I am going to use the FullDuplex example which is part of the samples provided by NServiceBus.
And let’s see how it looks from the visual studio.
And when we look into the project properties you can see that the project type is an output library and it is configured to run with an external program which is the NServiceBus.Host.exe.
So let’s run a typical failover / offline scenario and run the client application in isolation
and as you can see in the above diagram that the client can send messages even if the server is not running.
Let’s run the server application with debugging and mimic as if there is a failure. For example the finance system is down, exchange server in unavailable or any system to system integration failure. So in this scenario when we run the server application we see this exception thrown and seen like this in the console window.
Interesting thing is that if you have a break point in the server application set correctly you will see that the code gets retried 5 times before the exception is shown in the console.
This happens due to the fact that the server application is configured to use transactional queues and in the config file it has been configured to retry 5 times.
Upon close inspection of error queue this is what it looks like
And when we click to open the message you will see this is how the message looks like.
Next thing we have to do is fix the problem and whether the problem was due to hardware, software infrastructure failures let’s pretend those problems are resolved. So now what we have to do is put the message(s) back into where they belong and run our server component and there is exactly a tool called “ReturnedToSourceQueue” in the tools folder.Let’s run the tool.
This first thing you have to enter is the name of your error queue and after that you can specify a specific message with the Guid or all of it. For demonstration sake I’ll specify the keyword ‘all’ and you will see the message gets disappeared from the error queue and reappears in the server input queue.
and the message reappears in the input queue.
And now when we run our server component you will see the message gets processed.
As you can see we can fine tune and design these components to integrated different integration points and have that failover/durability scenario taken care by NServiceBus, we can take the message and transform into different type and let the handler of the new type take care of it.
For example our FinanceMessageHandler talks to the Finance system and processes the order if it fails to do so the order remains in the error. We don’t have to hack into some database or ask user to reenter the order. As soon as we resolved the Finance system failure/downtime the message will be processed successfully.On succesfully processing of this business process we can transform the message into an Email message and puts it into EmailQueue.
Now our EmailMessageHandler picks up the message and tries to process and if it encounters any problem with the mail server the message gets routed to the error queue that’s it. This is like creating a messaging eco-system where each system talks to each other in terms of a message and we are building this orchestration.This is what I meant by “Pass the Parcel (message)” pattern in my previous post.
That’s it for now and in my next post I am going to talk about the Generic host and what wonders it can create for your enterprise.
A couple of months ago I gave this NServiceBus integration presentation/walk through to my colleagues and thought about blogging it but for some unforeseen circumstances I couldn’t and today here I am with the presentation.
If you have no idea of what NServiceBus please read my previous post here.
Coming to the point straight lets start with an architecture diagram.
So the idea behind this architecture is that we will be building services which listens for a particular message and based on the message and it’s content it interact/performs an atomic operation and on completion forwards new message back into the queue.
For example in my project the “System A” is a finance system and as a result of an online financial transaction the “Message A” gets written to our Queue.As part of our NService infrastructure this message gets automatically picked by a handler, as in NService Universe for every message there has to be handler and each of these service interaction happens through a message.(We will see these handlers and message code in the second part)
Once our service has accomplished the task it simply writes a new message say “Message B” and it’s job is finished. In my case the “System B” is an enterprise wide emailing system and as soon as it gets an acknowledgement that Finance system operation is completed we are ready to send some emails and do some more interaction with other systems.
For me this is the core of designing integration system and what we are seeing here is that these services are autonomous and has no dependency on each other or the other systems.
For example if the online order submitted by a user is successful and the message gets written to the “Finance Queue” and at that time if the finance system is down still we don’t have any problems as the service will fail to perform its operation and the message will be sitting in the “Error Queue”, then as soon as Finance system is up and running we move the message from error queue to its processing queue and our job is done .
After this if for instance the Email system is down due routine updates , system failure etc still our message is sitting in the error queue and it’s going nowhere. This is exactly what one of the four tenets of Service Oriented Architecture is…
- Boundaries are explicit
- Services are autonomous
- Services share schema and contract, not class
- Service compatibility is based on policy
This is what I call as “pass the parcel” (Message) and MSMQ and NServiceBus will take care of all the infrastructure work. I don’t have to think about transactions failure and how to overcome them in case of catastrophic failures, as an architect my main job is to focus on business needs and how best to implement them.
This concludes the first part of it and I guess with this I have given you some context to the so-called “SOA” and how we can design our “SOA”, I know I didn’t cover anything substantial about NServiceBus but at times some architecture principle and theory are important before we can build the architecture.
Part 2 follow soon ….