Richard Seroter's Architecture Musings

Author: Richard Seroter

Considerations When Retrying Failed Messages in BizTalk or the ESB Toolkit
I was doing some research lately into a publish/subscribe scenario and it made me think of a “gotcha” that folks may not think about when building this type of messaging solution.

Specifically, what are the implications of resubmitted a failed transmission to a particular subscriber in a publish/subscribe scenario? For demonstration purposes, let’s say I’ve got a schema defining a purchase request for a stock.

Now let’s say that this is NOT an idempotent message and the subscriber only expects a single delivery. If I happen to send the above message twice, then 400 shares would get bought. So, we need a guaranteed-once delivery. Let’s also assume that we have multiple subscribers of this data who all do different things. In this demonstration, I have a single receive port/location which picks up this message, and two send ports which both subscribe on the message type and transmit the data to different locations.

As you’d expect, if I drop a single file in, I get two files out. Now what if the first send port fails for whatever reason? If I change the endpoint address to something invalid, the first port will fail, and the second will proceed as normal.

You can see that this suspension is directly associated with a particular send port, so resuming this failed message (after correcting the invalid endpoint address) should ONLY target the failed send port, and not put the message in a position to ALSO be processed by the previously-successful send port. This is verified in the scenario above.

So all is good. BUT what happens if you leverage an external system to facilitate the repair and resubmit of failed messages? This could be a SharePoint solution, custom application or the ESB Toolkit. Let’s use the ESB Toolkit here. I went into each send port and checked the Enable routing for failed messages box. This will result in port failures being published back to the bus where the ESB Toolkit “catch all” exception send port will pick it up.

Before testing this out, make sure you have an HTTP receive location set up. We’ll be using this to send message back from the ESB portal to BizTalk for reprocessing. I hadn’t set up an HTTP receive location yet on my IIS 7 box and found the instructions here (I used an HTTP receive location instead of the ESB on-ramps because I saw the same ESB Toolkit bug mentioned here).

So once again, I changed a send port’s address to something invalid and published a message to BizTalk. One message succeeded, one failed and there were no suspended messages because I had the failed message routing turned on. When I visit my ESB Toolkit Management Portal I can see the failed message in all its glory.

Clicking on the error drills into the details. From here I can view the message, click Edit and choose to resubmit it back to BizTalk.

This message comes back into BizTalk with no previous context or destination target. Rather, it’s as if I’m dropping this message into BizTalk for the first time. This means that ALL subscribers (in my scenario here) will get the message again and cause unintended side effects.

This is a case you may not think of when working primarily in point-to-point solutions. How do you get around it? A few ways I can think of:
- Build your messages and services to be idempotent. Who cares if a message comes once or ten times? Ideally there is a single identifier in each message that can indicate a message is a duplicate, or, the message itself is formatted in a way which is immune to retries. For instance, instead of the message saying to buy 200 shares, we could have fields with a “before amount” of 800 and “after amount” of 1000.
- Transform messages at the send port to destination specific formats. If each send port transforms the message to a destination format, then we could repair and resubmit it and only send ports looking for either the canonical format OR the destination format would pick it up.
- Have indicators in the message to indicate targets/retries and filter those out of send ports. We could add routing instructions to a message that specified a target system and have filters in send ports so only ports listening for that target pick up a message. The ESB Toolkit lets us edit the message itself before resubmitting it, so we could have a field called “target” and manually populate which send port the message should aim for.
So there you go. When working solely within BizTalk for messaging exceptions, the fact of using pub/sub or not shouldn’t matter. But, if you leverage error handling orchestrations or completely external exception management systems, you need to take into account the side effects of resubmitted messages that could reach multiple subscribers.
January 18, 2010
Interview Series: Four Questions With … Michael Stephenson

Happy New Year to you all! This is the 16th interview in my series of chats with thought leaders in the “connected systems” space. This month we have the pleasure of harassing Michael Stephenson who is a BizTalk MVP, active blogger, independent consultant, user group chairman, and secret lover of large American breakfasts.

Q: You head up the UK SOA/BPM User Group (and I’m looking forward to my invitation to speak there). What are the topics that generate the most interest, and what future topics do you think are most relevant to your audience?

A: Firstly, yes we would love you to speak, and ill drop you an email so we can discuss this 🙂

The user group actually formed about 18 months ago when two groups of people got together. There was the original BizTalk User Group and some people who were looking at a potential user group based around SOA. The people involved were really looking at this from a Microsoft angle so we ended up with the UK SOA/BPM User Group (aka SBUG). The idea behind the user group is that we would look at things from an architecture and developer perspective and be interested in the technologies which make up the Microsoft BPM suite (including ISV partners) and the concepts and ideas which go with solutions based on SOA and BPM principles.

We wanted to have a number of themes going on and to follow some of the new technologies coming out which organizations would be looking at. Some of the most common technology topics we have had previously have included BizTalk, Dublin, Geneva and cloud. We have also tried to have some ISV sessions too. My idea around the ISV sessions is that most people tend to see ISV’s present high level topics at big industry events where you see pretty slides and quite simple demonstrations but with the user group we want to give people the change to get a deeper understanding of ISV offerings so they know how various products are positioned and what they offer. Some examples we have coming up on this front are in January where Global 360 will be doing a case study around Nationwide Building Society in the UK and AgilePoint will be doing a web cast about SAP. Hopefully members get a change to see what these products do, and to network and ask tough questions without it being a sales based arena.

Last year one of our most popular sessions was when Darren Jefford joined us to do a follow up to a session he presented at the SOA/BPM Road show about on-premise integration to the cloud. I’m hoping that Darren might be able to join us again this year to do another follow up to a session he did recently about a BizTalk implementation with really high performance characteristics. Hopefully the dates will workout well for this.

We have about 4 in person meetings per year at the moment, and a number of online web casts. I think we have got things about right in terms of technology sessions, and I expect that in the following year we will combine potentially BizTalk 2009 R2, and AppFabric real world scenarios, more cloud/Azure, and I’d really like to involve some SharePoint stuff too. I think one of the weaker areas is around the concepts and ideas of SOA or BPM. I’d love to get some people involved who would like to speak about these things but at present I haven’t really made the right contacts to find appropriate speakers. Hopefully this year we will make some inroads on this. (Any offers please contact me).

A couple of interesting topics in relation to the user group are probably SQL Server, Oslo and Windows Workflow. To start with Windows Workflow is one of those core technologies which you would expect the technology side of our user group to be pretty interested in, but in reality there has never been that much appetite for sessions based around WF and there hasn’t really been that many interesting sessions around it. You often see things like here is how to do a work flow that does a specific thing, but I haven’t really seen many cool business solutions or implementations which have used WF directly. I think the stuff we have covered previously has really been around products which leverage workflow. I think this will continue but I expect as AppFabric and a solid hosting solution for WF becomes available there may be future scenarios where we might do case studies of real business problems solved effectively using WF and Dublin.

Oslo is an interesting one for our user group. Initially there was strong interest in this topic and Robert Hogg from Black Marble did an excellent session right at the start of our user group about what Oslo was and how he could see it progressing. Admittedly I haven’t been following Oslo that much recently but I think it is something I will need to get feedback from our members to see how we would like to continue following its development. Initially it was pitched as something which would definitely be of interest to the kind of people who would be interested in SBUG but since it has been swallowed up by the “SQL Server takes over the world” initiative, we probably need to just see how this develops, certainly the core ideas of Oslo still seem to be there. SQL Server also has a few other features now such as StreamInsight which are probably also of interest to SBUG members.

I think one of the challenges for SBUG in the next year is about the scope of the user group. The number of technologies which are likely to be of interest to our members has grown and we would like to get some non technology sessions involved also, so the challenge is how we manage this to ensure that there is a strong enough common interest to keep involved, yet the scope should be wide enough to offer variety and new ideas.

If you would like to know more about SBUG please check out our new website on: http://uksoabpm.org.

Q: You’ve written a lot on your blog about testing and management of BizTalk solutions. In your experience, what are the biggest mistakes people make when testing system integration solutions and how do those mistakes impact the solution later on?

A: When it comes to BizTalk (or most technology) solutions there are often many ways to solve a problem and produce a solution that will do a job for your customer to one degree or another. A bad solution can often still kind of work. However when it comes to development and testing processes it doesn’t matter how good your code/solution is if the process you use is poor, you will often fail or make your customer very angry and spend a lot of their money. I’ve also felt that there has been plenty of room for blogging content to help people with this. Some of my thoughts on common mistakes are:

Not Automating Testing

This can be the first step to making your life so much less stressful. On the current project I’m involved with we have a large number of separate BizTalk applications each with quite different requirements. The fact that all of these are quite extensively tested with BizUnit means that we have quite low maintenance costs associated with these solutions. Anytime we need to make changes we always have a high level of confidence that things will work well.

I think on this project during its life cycle the defects associated with our team have usually been <5% related to coding errors. The majority are actually because external UAT or System test teams have written tests incorrectly, problems with other systems which get highlighted by BizTalk or a poor requirement.

Good automated testing means you can be really proactive when it comes to dealing with change and people will have confidence in the quality of things you produce.

Not Stubbing Out Dependencies

I see this quite often when you have multiple teams working on a large development project. Often the work produced by these teams will require services from other applications or a service bus. So many times I’ve seen the scenario where the developer on Team A downs tools because their code wont work because the developer on Team B is making changes to the code which runs on his machine. In the short term this can cause delays to a project, and in the longer term a maintenance nightmare. When you work on a BizTalk project you often have this challenge and usually stubbing out these dependencies becomes second nature. Sometimes its the teams who don’t have to deal with integration regularly who aren’t used to this mindset.

This can be easily mitigated if you get into the contract first mind set and its easy to create a stub of most systems that use a standards based interface such as web services. I’d recommend checking out Mockingbird as one tool which can help you here. Actually to plug SBUG again we did a session about Mockingbird a few months ago which is available for download: http://uksoabpm.org/OnlineMiniMeetings.aspx

Not considering data flow across systems

One common bad practice I see when someone has automated testing is that they really just check the process flow but don’t really consider the content of messages as they flow across systems. I once saw a scenario where a process passed messages through BizTalk and into an internal LOB system. The development team had implemented some tests which did pretty good job at testing the process, but the end to end system testing was performed by an external testing team. This team basically loaded approximately 50k messages per day for months through the system into the LOB application and made a large assumption that because there were no errors recorded by the LOB application everything was fine.

It turned out that a number of the data fields were handled incorrectly by the LOB application and this just wasn’t spotted.

The lessons here were mainly that sometimes testing is performed by specialist testing teams and you should try develop a relationship between your development and test teams so you know what everyone is doing. Secondly executing millions of messages is no where near as effective as understanding the real data scenarios and testing those.

Poor/No/Late Performance Testing

This is one of the biggest risks of any project and we all know its bad. Its not uncommon for factors beyond our control to limit our ability to do adequate performance testing. In BizTalk world we often have the challenge that test environments do not really look like a production environment due to the different scaling options taken.

If you find yourself in this situation probably the best thing you can do is to firstly ensure the risk is logged and that people are aware of the risk. If your project has accepted the risk and doesn’t plan to do anything about it, the next thing is to agree as a team how you will handle this. Agree a process of how you will ensure to maximize the resources you do have to adequately performance test your solution. Maybe this is to run some automated tests using BizUnit and LoadGen on a daily basis, maybe its to ensure you are doing some profiling etc. If you agree your process and stick to it then you have mitigated the risk as much as possible.

A couple of additional side thoughts here are that a good investment in the monitoring side of your solution can really help. If you can see that part of your solution isn’t performing too well in a small test environment don’t just disregard this because the environment is not production like, analyze the characteristics of the performance and understand if you can make optimizations. The final thought here is that when looking at end to end performance you also need to consider the systems you will integrate with. In most scenarios latency or throughput limitations of an application you integrate with will become a problem before any additional overhead added by BizTalk.

Q: When architecting BizTalk solutions, you often make the tradeoff between something that is either (a) quite complex, decoupled and easier to scale and change, or (b) something a bit more rigid but simpler to build, deploy and maintain. How do you find the right balance between those extremes and deliver a project on time and architected the “right” way for the customer?

A: By their nature integration projects can be really varied, and even seasoned veterans will come across scenarios which they haven’t seen before or a problem with many ways to solve it. I think its very helpful if you can be open-minded and able to step back and look at the problem from a number of angles, consider the solution from the perspective of all of your stakeholders. This should help you to evaluate the various options. Also one of my favorite things to do is to bounce the idea of some friends. You often see this on various news groups or email forums. I think sometimes people are afraid to do this, but you know, no one knows everything and people on these forums generally like to help each other out so its a very valuable resource to be able to bounce your thoughts off colleagues (especially if your project is small).

More specifically about Richard’s question I guess there is probably two camps on this, the first is “Keep it simple stupid”, and as a general rule if you do what you are required to do, do it well and do it cheaply then usually everyone will be happy. The problem with this comes when you can see there are things past the initial requirements which you should consider now or the longer term cost will be significantly higher. The one place you don’t want to go is where you end up lost in a world of your own complexity. I can think of a few occasions where I have seen solutions where the design had been taken to the complex extreme. While designing or coding, if you can teach yourself to regularly take a step away from your work and ask yourself “What is it that I’m trying to do” or to explain things to a colleague you will be surprised how many times you can save yourself a lot of headaches later.

I think one of the real strengths of BizTalk as a product is that it lets you have a lot of this flexibility without too much work compared to non BizTalk based approaches. I think in the current economic climate it is more difficult to convince a customer about the more complex decoupled approaches when they cant clearly and immediately see benefits from it. Most organizations are interested in cost and often the simpler solution is perceived to be the cheapest. The reality is that because BizTalk has things like the pub/sub model, BRE, ESB Guidance, etc it means you can deal with complexity and decoupling and scaling without it actually getting too complex. To give you a recent and simple example of this, one of my customers wanted to have a quick and simple way of publishing some events to a B2B partner from a LOB application. Without going into too much detail this was really easy to do, but the fact that it was based on BizTalk meant the decoupling offered by subscriptions allowed us to reuse this process three more times to publish events to different business partners in different formats over different protocols. This was something the customer hadn’t even thought about initially.

I think on this question there is also the risk factor to consider, when you go for the more complex solution the perceived risk of things going wrong is higher which tends to turn some people away from the approach, however this is where we go back to the earlier question about testing and development processes. If you can be confident in delivering something which is of high quality then you can be more confident in delivering something which is more complex.

Q [stupid question]: As we finish up the holiday season, I get my yearly reminder that I am utterly and completely incompetent at wrapping gifts. I usually end these nightmarish sessions completely hairless and missing a pint of blood. What is an example of something you can do, but are terrible at, and how can you correct this personal flaw?

A: I feel your pain on the gift wrapping front (literally). I guess anyone who has read this far will appreciate one of my flaws is that I can go on a bit, hope some of it was interesting enough!

I think the things that I like to think I can do, but in reality I’d have to admit I am terrible at are Cooking and DIY. Both are easily corrected by getting other people to do them, but saying as this will be the first interview of the new year I guess its fitting that I should make a new years resolution so I’ll plan to do something about one of them. Maybe take a cooking class.

Oh did I mention another flaw is that I’m not too good at keeping new years resolutions.

Thanks to Mike for taking the time to entertain us and provide some great insights.

January 3, 2010
MVP Again

Thanks to the Microsoft team for granting me a third straight BizTalk MVP. I suspect that they keep doing this just to see what regrettable statements I’ll make at the next MVP Summit. Either way, it’s an honor to receive, and I’m grateful for the perks it has to offer.

I need to get ready for the Summit in February by preparing my list of ludicrous product features which I’ll be demanding be added to BizTalk Server 2011 (e.g. use text files for storage instead of the MessageBox, XBox-themed skins for the Mapper, functoids that multiply any number by pi, low latency processing, etc).

Anyway, thanks to all of you who continue to visit here, buy my book, and put up with my shenanigans. Always appreciated.

January 2, 2010
Populating Word 2007 Templates Through Open XML
I recently had a client at work interested in populating contracts out of the information stored in their task tracking tool. Today this is a manual process where the user opens up a Microsoft Word template and retypes the data points stored in their primary application.

I first looked at a few commercial options, and then got some recommendations from Microsoft to look deeper into the Open XML SDK and leverage the native XML formats of the Office 2007 document types. I found a few articles and blog posts that explained some of the steps, but didn’t seem to find a single source of the whole end to end process. So, I figured that I’d demonstrate the prototype solution that I built.

First, we need a Word 2007 document. Because I’m not feeling particular frisky today, I will fill this document with random text using the ever-useful “=rand(9)” command to make Word put 9 random paragraphs into my document.

Next, I switch to the Developer tab to find the Content Controls I want to inject into my document. Don’t see the Developer tab? Go here to see how to enable it.

Now I’m going to sprinkle a few Text Content Controls throughout my document. The text of each control should indicate the type of content that goes there. For each content control on the page, select it and choose the Properties button on the ribbon so that you can provide the control with a friendly name.

At this point, I have four Content Controls in my document and each has a friendly title. Now we can save and close the document. As you probably know by now, the Office 2007 document formats are really just zip files. If you change the extension of our just-saved Word doc from .docx to .zip, you can see the fun inside.

I looked a few options for manipulating the underlying XML content and finally ended up on the easiest way to update my Content Controls with data from outside. First, download the Word 2007 Content Control Toolkit from CodePlex. Then install and launch the application. After browsing to our Word document, we see our friendly-named Content Controls in the list.

You’ll notice that the XPath column is empty. What we need to do next is define a Custom XML Part for this Word document, and tie the individual XML nodes to each Content Control. On the right hand side of the Word 2007 Content Control Toolkit you’ll see a window that tells us that there are currently no custom XML parts in the document.

The astute among you may now guess that I will click the “Click here to create a new one.” I have smart readers. After choosing to create a new part, I switched to the Edit view so that I could easily hand craft an XML data structure.

For a more complex structure, I could have also uploaded an existing XML structure. The values I put inside each XML node are the values that the Word document will display in each content control. Switch to the Bind view and you should see a tree structure.

Click each node, and then drag it to the corresponding Content Control. When all four are complete, the XPath column in the Content Controls should be populated.

Go ahead and save the settings and close the tool. Now, if we once again peek inside our Word doc by changing it’s extension to .zip, we’ll see a new folder called CustomXml that has our XML definition in there.

For my real prototype I built a WCF service that created the Word documents out of the templates and loaded them into SharePoint. For this blog post, I’ll resort to a Console application which reads the template and emits the resulting Word document to my Desktop. You’ll get the general idea though.

If you haven’t done so already, download and install the Open XML Format SDK 1.0 from Microsoft. After you’ve done that, create a new VS.NET Console project and add a reference to DocumentFormat.OpenXML. Mine was found here: C:\Program Files\OpenXMLSDK\1.0.1825\lib\DocumentFormat.OpenXml.dll. I then added the following “using” statements to my console class.
```
using DocumentFormat.OpenXml;
using DocumentFormat.OpenXml.Packaging;
using System.Xml; using System.IO;
```
Next I have all the code which makes a copy of my template, loads up the Word document, removes the existing XML part, and adds a new one which has been populated with the values I want within the Content Controls.
```
static void Main(string[] args)
        {

            Console.WriteLine("Starting up Word template updater ...");

            //get path to template and instance output
            string docTemplatePath = @"C:\Users\rseroter\Desktop\ContractSample.docx";
            string docOutputPath = @"C:\Users\rseroter\Desktop\ContractSample_Instance.docx";

            //create copy of template so that we don't overwrite it
            File.Copy(docTemplatePath , docOutputPath);

            Console.WriteLine("Created copy of template ...");

            //stand up object that reads the Word doc package
            using (WordprocessingDocument doc = WordprocessingDocument.Open(docOutputPath, true))
            {
                //create XML string matching custom XML part
                string newXml = "<root>" +
                    "<Location>Outer Space</Location>" +
                    "<DocType>Contract</DocType>" +
                    "<MenuOption>Start</MenuOption>" +
                    "<GalleryName>Photos</GalleryName>" +
                    "</root>";

                MainDocumentPart main = doc.MainDocumentPart;
                main.DeleteParts<CustomXmlPart>(main.CustomXmlParts);

                //add and write new XML part
                CustomXmlPart customXml = main.AddNewPart<CustomXmlPart>();
                using (StreamWriter ts = new StreamWriter(customXml.GetStream()))
                {

                    ts.Write(newXml);
                }

            //closing WordprocessingDocument automatically saves the document
            }

            Console.WriteLine("Done");
            Console.ReadLine();
        }
```
When I run the console application, I can see a new file added to my Desktop, and when I open it, I find that my Content Controls now have the values that I set from within my Console application.

Not bad. So, as you can imagine, it’s pretty simple to now take this Console app, and turn it into a service which takes in an object containing the data points we want added to our document. So while this is hardly a replacement for a rich content management or contract authoring tool, it is a quick and easy way to do a programmatic mail merge and update existing documents. Heck, you could even call this from a BizTalk application or custom application to generate documents based on message payloads. Fun stuff.
December 23, 2009
Building WCF Workflow Services and Hosting in AppFabric
Yesterday I showed how to deploy the new WCF 4.0 Routing Service within IIS 7. Today, I’m looking at how to take one of those underlying services we built and consume it from a WCF Workflow Service hosted in AppFabric.

In the previous post, I created a simple WCF service called “HelloServiceMan” which takes a name and spits back a greeting. In this post, I will use this service completely illogically and only to prove a point. Yes, I’m too lazy right now to create a new service which creates a more realistic scenario. What I DO want is to call into my workflow, immediately send a response back, and then go about calling my existing web service. I’m doing this to show that if my downstream service was down, my workflow (hosted with AppFabric) can be suspended, and then resume once my downstream service comes back online. Got it? Cool.

First, we need a WCF Workflow Service app. In VS 2010, I pick this from the “Workflow” section.

I then added a single class file to this project which holds data contracts for the input and output message of the workflow service.
```
[DataContract(Namespace="https://seroter.com/Contracts")]
   public class NewOrderRequest
   {
       [DataMember]
       public string ProductId { get; set; }
       [DataMember]
       public string CustomerName { get; set; }
   }

   [DataContract(Namespace = "https://seroter.com/Contracts")]
   public class OrderAckResponse
   {
       [DataMember]
       public string OrderId { get; set; }
   }
```
Next I added a Service Reference to my existing WCF service. This is the one that I plan to call from within the workflow service. Once I have my reference defined, and build my project, a custom Workflow Activity should get added to my Toolbox.

If you’re familiar with building BizTalk orchestrations, then working with the Windows Workflow design interface is fairly intuitive. Much like an orchestration, the first thing I do here is define my variables. This includes the default “correlation handle” object which was already there, and then variables representing the input/output of my workflow service, and the request/response messages of my service reference.

Notice that for variables which aren’t explicitly instantiated by receiving messages into the workflow (i.e. initial received message, response from service call) have explicit instantiation in the “Default” column.

Next I sketched out the first part of the workflow which receives the inbound “order request” (defined in the above data contract), sets a tracking number and returns that value to the caller. Think of when you order a package from an online merchant and they immediately ship you a tracking code while starting their order processing behind the scenes.

Next I call my referenced service by first setting the input variable attribute value, and then using the custom Workflow Activity shape which encapsulates the service request and response (once again, realize that this content of this solution makes no sense, but the principles do).

After building the solution successfully, we can get this deployed to IIS 7 and running in the AppFabric. After creating an IIS web application which points to this solution, we can right click our new application and choose .NET 4 WCF and WF and then Configure.

On the Workflow Persistence tab, I clicked the Advanced button and made sure that on unhandled errors that I abandon and suspended.

If you are particularly astute, you may notice at the top of the previous image that there’s an error complaining about the net.pipe protocol missing from my Enabled Protocols. HOWEVER, there is a bug/feature in this current release where you should ignore this and ONLY add net.pipe to the Enabled Protocols at the root web site. If you put it down at the application level, you get bad things.

So, now I can browse to my workflow service and see a valid service endpoint.

I can call this service from the WCF Test Client, and hopefully I not only get back the immediate response, but also see a successfully completed workflow in the AppFabric console. Note that if you don’t see things showing up in your AppFabric console, check your list of Windows Services and make the sure the Application Server Event Collector is started.

Now, let’s turn off the WCF service application so that our workflow service can’t complete successfully. After calling the service again, I should still get an immediate response back from my workflow since the response to the caller happens BEFORE the call to the downstream service. If I check the AppFabric console now, I see this:

What the what?? The workflow didn’t suspend, and it’s in a non-recoverable state. That’s not good for anybody. What’s missing is that I never injected a persistence point into my workflow, so it doesn’t have a place to pick up and resume. The quickest way to fix this is to go back to my workflow, and on the response to the initial request, set the PersistBeforeSend flag so that the workflow forces a persistence point.

After rebuilding the service, and once again shutting down the downstream service, I called my workflow service and got this in my AppFabric console:

Score! I now have a suspended instance. After starting my downstream service back up, I can select my suspended instance and resume it.

After resuming the instance, it disappears and goes under the “Completed Instances” bucket.

There you go. For some reason, I just couldn’t find many examples at all of someone building/hosting/suspending WF 4.0 workflow services. I know it’s new stuff, but I would have thought there was more out there. Either way, I learned a few things and now that I’ve done it, it seems simple. A few days ago, not so much.
December 17, 2009
Hosting the WCF 4.0 Routing Service in IIS 7
I recently had occasion to explore the new WCF 4.0 Routing Service and thought I’d share how I set up a simple solution that demonstrated its capabilities and highlights how to host it within IIS.

[UPDATE: I’ve got a simpler way to do this in a later post that you can find here.]

This new built-in service allows us to put a simple broker in front of our services and route inbound messages based on content, headers, and more. Problem for me was that every demo I’ve seen of this thing (from PDC, and other places) show simple console hosts for the service and not a more realistic web server host. This is where I come in.

First off, I need to construct the services that will be fronted by the Routing Service. In this simple case, I have two services that implement the same contract. In essence, these services take a name and gender, and spit back the appropriate “hello.” The service and data contracts look like this:
```
[ServiceContract]
    public interface IHelloService
    {
        [OperationContract]
        string SayHello(Person p);
    }

    [DataContract]
    public class Person
    {
        private string name;
        private string gender;

        [DataMember]
        public string Name
        {
            get { return name; }
            set { name = value; }
        }

        [DataMember]
        public string Gender
        {
            get { return gender; }
            set { gender = value; }
        }
    }
```
I then have a “HelloServiceMan” service and “HelloServiceWoman” service which implement this contract.
```
public class HelloServiceMan : IHelloService
    {

        public string SayHello(Person p)
        {
            return "Hey Mr. " + p.Name;
        }
    }
```
I’ve leveraged the new default binding capabilities in WCF 4.0 and left my web.config file virtually empty. After deploying these services to IIS 7.0, I can use the WCF Test Client to prove that the service performs as expected.

Nice. So now I can add the Routing Service. Now, what initially perplexed me is that since the Routing Service is self contained, you don’t really have a *.svc file, but then I didn’t know how to build a web project that could host the service. Thanks to Stephen Thomas (who got code from the great Christian Weyer) I got things working.

You need three total components to get this going. First, I created a new, Empty ASP.NET Web Application project and added a .NET class file. This class defines a new ServiceHostFactory class that the Routing Service will use. That class looks like this:
```
class CustomServiceHostFactory : ServiceHostFactory
{
    protected override System.ServiceModel.ServiceHost CreateServiceHost(System.Type serviceType, System.Uri[] baseAddresses)
    {
        var host = base.CreateServiceHost(serviceType, baseAddresses);

        var aspnet = host.Description.Behaviors.Find<AspNetCompatibilityRequirementsAttribute>();

        if (aspnet == null)
        {
            aspnet = new AspNetCompatibilityRequirementsAttribute();
            host.Description.Behaviors.Add(aspnet);
        }

        aspnet.RequirementsMode = AspNetCompatibilityRequirementsMode.Allowed;

        return host;
    }
}
```
Here comes the tricky, but totally logical part. How do you get the WCF Routing Service instantiated? Add a global.asax file to the project and add the following code to the Application_Start method:
```
using System.ServiceModel.Activation;
using System.ServiceModel.Routing;
using System.Web.Routing;

namespace WebRoutingService
{
    public class Global : System.Web.HttpApplication
    {

        protected void Application_Start(object sender, EventArgs e)
        {
            RouteTable.Routes.Add(
               new ServiceRoute("router", new CustomServiceHostFactory(),
                   typeof(RoutingService)));
        }
```
Here we stand up the Routing Service with a “router” URL extension. Nice. The final piece is the web.config file. Here is where you actually define the Routing Service relationships and filters. Within the system.serviceModel tags, I defined my client endpoints that the router can call.
```
<client>
      <endpoint address="http://localhost/FirstWcfService/HelloServiceMan.svc"
          binding="basicHttpBinding" bindingConfiguration="" contract="*"
          name="HelloMan" />
      <endpoint address="http://localhost/FirstWcfService/HelloServiceWoman.svc"
          binding="basicHttpBinding" bindingConfiguration="" contract="*"
          name="HelloWoman" />
    </client>
```
The Routing Service ASP.NET project does NOT have any references to the actual endpoint services, and you can see here that I ignore the implementation contract. The router knows as little as possible about the actual endpoints besides the binding and address.

Next we have the brand new “routing” configuration type which identifies the filters used to route the service messages.
```
<routing>
      <namespaceTable>
        <add prefix="custom" namespace="http://schemas.datacontract.org/2004/07/FirstWcfService"/>
      </namespaceTable>
      <filters>
        <filter name="ManFilter" filterType="XPath" filterData="//custom:Gender = 'Male'"/>
        <filter name="WomanFilter" filterType="XPath" filterData="//custom:Gender = 'Female'"/>
      </filters>
      <filterTables>
        <filterTable name="filterTable1">
          <add filterName="ManFilter" endpointName="HelloMan" priority="0"/>
          <add filterName="WomanFilter" endpointName="HelloWoman" priority="0"/>
        </filterTable>
      </filterTables>
    </routing>
```
I first added a namespace prefix table, then have a filter collection which, in this case, uses XPath against the inbound message to determine the gender value within the request. Note that if you want to use a comparison operation such as “<” or “>”, you’ll have to escape it in this string to “>” or “<”. Finally, I have a filter table which maps a particular filter to which endpoint should be applied.

Finally, I have the service definition and behavior definition. These both leverage objects and configuration items new to WCF 4.0. Notice that I’m using the “IRequestReplyRouter” contract since I have a request/reply service being fronted by the Routing Service.
```
<services>
      <service behaviorConfiguration="RoutingBehavior" name="System.ServiceModel.Routing.RoutingService">
        <endpoint address="" binding="basicHttpBinding" bindingConfiguration=""
          name="RouterEndpoint1" contract="System.ServiceModel.Routing.IRequestReplyRouter" />
      </service>
    </services>
    <behaviors>
      <serviceBehaviors>
        <behavior name="RoutingBehavior">
          <routing routeOnHeadersOnly="false" filterTableName="filterTable1" />
          <serviceDebug includeExceptionDetailInFaults="true"/>
          <serviceMetadata httpGetEnabled="true" />
        </behavior>
      </serviceBehaviors>
    </behaviors>
```
Once we build and deploy the service to IIS 7, we can browse it. Recall that in our global.asax file we defined a URL suffix named “router.” So, to hit the service, we load our web application and append “router.”

As you’d expect, this WSDL tells us virtually nothing about what data this service accepts. What you can do from this point is build a service client which points at one of the actual services (e.g. “HelloServiceMan”), but then switch the URL address in the application’s configuration file. This way, you can still import all the necessary contract definitions, while then switching to leverage the content-based routing service.

So, the Routing Service is pretty cool. It does a light-weight version of what BizTalk does for routing. I haven’t played with composite filters and don’t even know if it’s possible to have multiple filter criteria (like you can with a BizTalk Server subscription). Either way, it’s good to know how to actually deploy this new capability in an enterprise web server instead of a console host.

Anyone else have lessons learned with the Routing Service?
December 16, 2009
Interview Series: Four Questions With … Brian Loesgen

Happy December and welcome to my 15th interview with a leader in the “connected technology” space. This month, we sit down with Brian Loesgen who is a prolific author, blogger, speaker, salsa dancer, former BizTalk MVP, and currently an SOA architect with Microsoft.

Q: PDC 2009 has recently finished up and we saw the formal announcements around Azure, AppFabric and the BizTalk Server roadmap. It seems we’ve been talking to death about BizTalk vs. Dublin (aka AppFabric) and such, so instead, talk to us about NEW scenarios that you see these technologies enabling for customers. What can Azure and/or AppFabric add to BizTalk Server and allow architects to solve problems easier than before?

A: First off, let me clarify that Windows Server AppFabric is not just “Dublin”-renamed, it brings together the technologies that were being developed as code-name “Dublin” and code-name “Velocity”. For the benefit of your readers that may not know much about “Velocity”, it was a distributed in-memory cache, ultra-high performance and highly-scalable. Although I have not been involved with “Velocity”, I have been quite close to “Dublin” since the beginning.

I think the immediate value people will see in Windows Server AppFabric is that now.NET developers are being provided with a host for their WF workflows. Previously developers could use SharePoint as a host, or write their own host (typically a Windows service). However, writing a host is a non-trivial task once you start to think about scale-out, failover, tracking, etc. I believe that the lack of a host was a bit of an adoption blocker for WF and we’re going to see that a lot of people that never really thought about writing workflows will start doing so. People will realize that a lot of what they write actually is a workflow, and that we’ll see migration once they see how easy it is to create a workflow and host it in AppFabric and expose it as WCF Web services. This doesn’t, of course, preclude the need for integration server (BizTalk) which as you’ve said we’ve been talked to death about, and “when to use what” is one of the most common questions I get. There is a time and place for each, they are complementary.

Your question is very astute, although Azure and AppFabric will allow us to create the same applications architected in a new way (in the cloud, or hybrid on-premise-plus-cloud), they will also allow us to create NEW types of applications that previously would not have been possible. In fact, I have already had many real-world discussions with customers around some novel application architectures.

For example, in one case, we’re working with geo-distributed (federated) ESBs, and potentially tens of thousands of data collection points scattered around the globe, each rolling up data to its “regional ESB”. Some of those connection points will be in VERY remote locations, where reliable connectivity can be a problem. It would never have been reasonable to assume that those locations would be able to establish secure connections to a data center with any kind of tolerable latency, however, the expectation is that somehow they’ll all be able to reach to cloud. As such, we can use the Windows Azure platform Service Bus as a data collection and relay mechanism.

Another cool pattern is using the Windows Azure platform Service Bus as an entry point into an on-premises ESB. In the past, if a BizTalk developer wanted to accept input from the outside typically they would expose a Web service, and reverse-proxy that to make it available, probably with a load balancer thrown in if there was any kind of sustained or spike volume. That all works, but it’s a lot of moving parts that need to be set up. A new pattern we can do now is that we can use the Windows Azure platform Service Bus as a relay: externals parties send messages to it (assuming they are authorized to do so by the Windows Azure platform Access Control service) , and a BizTalk receive location picks up from it. That receive location could even be an ESB on-ramp. I have a simple application that integrates BizTalk, ESB, BAM, SharePoint, InfoPath, SS/AS, SS/RS (and a few more things). It was trivial for me to add another receive location that picked up from the Service Bus (blog post about this coming soon, really J). Taking advantage of Azure to act as an intermediary like this is a very powerful capability, and one I think will be very widely used.

Q: You were instrumental in incubating and delivering the first release of the ESB Toolkit (then ESB Guidance) for Microsoft. I’ve spoken a bit about itineraries here, but give me your best sales pitch on why itineraries matter to Microsoft developers who may not familiar with this classic ESB pattern. When is the right time to use them, why use an itinerary instead of an orchestration, when should I NOT use them?

A: I recently had the pleasure of creating and delivering a “Building the Modern ESB” presentation with a gentleman from Sun at the SOA Symposium in Rotterdam. It was quite enlightening, to see the convergence and how similar the approaches are. In that world, an itinerary is called a “micro-flow”, and it exists for exactly the same reasons we have it in the ESB Toolkit.

Itineraries can be thought of as a light-weight service composition model. If all you’re doing is receiving a request, calling 3 services, perhaps with some transformations along the way, and then returning a response, then that would be appropriate for an itinerary. However, if you have a more complex flow, perhaps where you require sophisticated branching logic or compensation, or if it’s long running, then this is where orchestrations come into play. Orchestration provides the rich semantics and constructs you need to handle these more sophisticated workflows.

The ESB Toolkit 2.0 added the new Broker capabilities, which allow you to do conditional branching from inside an itinerary, however I generally caution against that as its one more place you could hide business logic (although there will of course be times where this is the right approach to take).

A pattern that I REALLY like is to use BizTalk’s Business Rules Engine (BRE) to dynamically select an itinerary and apply it to a message. The BRE has always been a great way to abstract business rules (which could change frequently) from the process (which doesn’t change often). By letting the BRE choose an itinerary, you have yet another way to leverage this abstraction, and yet another way you can quickly respond to changing business requirements.

Q: You’ve had an interesting IT career with a number of strategic turns along the way. We’ve seen the employment market for developers change over the past few years to the point that I’ve had colleagues say that they wouldn’t recommend that their kids go into computer science and rather focus on something else. I still think that this is a fun field with compelling opportunities, but that we all have to be more proactive about our careers and more tuned in to the skills we bring to the table. What advice would you give to those starting out their IT careers with regards to where to focus their learning and the types of roles they should be looking for?

A: It’s funny, when the Internet bubble burst, a couple of developers I knew then decided to abandon the industry and try their hands at something else: being a mortgage broker. I’m not sure where they are now, probably looking for the next bubble J

You’re right though, I’ve been fortunate in that I’ve had the opportunity to play many roles in this industry, and I have never been as excited about where we are as an industry as I am now. The maturing and broad adoption of Web service standards, the adoption of Service-Oriented Architectures and ESBs, the rapid onset of the cloud (and new types of applications cloud computing enables)… these are all major change agents that are shaking up our world. There is, and will continue to be, a strong market for developers. However, increasingly, it’s not so much what you know, but how quickly you can learn and apply something new that really matters. In order to succeed, you need to be able to learn and adapt quickly, because our industry seems to be in a constant state of increasingly rapid change. In my opinion, good developers should also aspire to “move up the stack” if you want to advance your career, either along a technical track (as an architect) or perhaps a more business-related track (strategy advisor, project management). As you can see from the recently published SOA Manifesto (http://soa-manifesto.org), providing business value and better alignment between IT and business are key values that should be on the minds of developers and architects, and SOA “done right” facilitates that alignment.

Q [stupid question]: year you finally bit the bullet and joined Microsoft. This is an act often equated with “joining the dark side.” Now that phrase isn’t only equated with doing something purely evil, but rather, giving into something that is overwhelming and seems inevitable such as joining Facebook, buying an iPhone, or killing an obnoxious neighbor and stuffing him into a recycling container. Standard stuff. Give us an example (in life or technology) of something you’ve been resisting in your life, and why you’re holding out.

A: Wow, interesting question.

I would have to say Twitter. I resisted for a long time. I mean c’mon, I was already on Facebook, Live, LinkedIn, Plaxo…. Where does it end? At some point you need to be able to live your life and do your work rather than talking about it. Plus, maybe I’m verbose, but when I have something to say it’s usually more than 140 characters, so I really didn’t see the point. However, I succumbed to the peer pressure, and now, yes, I am on Twitter.

Do you think if I say something like “you can follow me at http://twitter.com/BrianLoesgen” that maybe I’ll get the “Seroter Bump”? 🙂

Thanks Brian for some good insight into new technologies.

December 3, 2009
Using UML vs. ADL (Arbitrary Diagramming Language)
At my company, we highly encourage our architects and developers to communicate their design models through UML. We have a standard set of diagrams that we regularly use to convey aspects of a given solution, including:
- Component diagrams for system dependency, functional decomposition and data flow
- Use case diagrams for business and system scenarios
- Sequence diagrams for a timing perspective on system interactions
- Deployment diagrams for both logical and physical representations of the hardware/software landscape
We all know that the job of architect pretty much consists of “ask lots of annoying questions” and “draw boxes and lines.” The downside with being so UML-centric is that UML can accidentally become the default way to draw boxes and lines that represent all sorts of things. My boss has encouraged us to be observant for when a particular graphical representation should take a more “marketecture” approach (i.e. fancy boxes and lines in PowerPoint or Visio) vs. formal modeling in UML or BPMN. She labeled the PowerPoint/Visio approach as using the Arbitrary Diagramming Language (ADL) method. Seems about right.

What’s an example of this? I’ll give you two. I recently had to put together a reference architecture for a project I’m working on. My first inclination was to throw together a functional decomposition diagram (component diagram) of all the relevant functional pieces. However, this was very early in the project lifecycle, and I knew that this diagram would be consistently referred to in both business and technical meetings. So instead of going nuts on a UML diagram that would be utterly confusing, and potentially hacked up to represent certain ideas, I went with a PowerPoint model that has been in regular use for 5 months now.

In another case, I’m trying to represent deployment options (incremental phased approach vs. all-at-once) for a global system. I began this effort thinking about component diagrams, deployment diagrams, swim lanes and mashing some things together. Again though, I’d probably have to bastardize UML and overload agreed-upon constructs in order to convey my point. So, I went again to trusty PowerPoint and modeled out the options in a more business-friendly way. I could create my own terminology/model without feeling guilty for screwing up standard UML.

This way, I could convey a sequence of deployment phases (through a set of slides), which parties were involved, and the necessary interim interfaces without butchering UML.

So when to use each modeling style? I’d like your thoughts as well, but for me, I switch to an ADL / marketecture model when:
- Doing early project stage diagrams to convey concepts presented to broad audiences (e.g. logical architectures, deployment diagrams, proposed data flow between systems/organizations)
- I realize that I need to cannibalize, bastardize, overload or otherwise distort UML notation to convey a point (e.g. user experience flow, etc)
I fire up a UML tool for modeling when:
- I’m building lasting representations of how a system will be built and providing an accurate blueprint to the development team
- Conveying solution aspects to solely technical teams through presentations or design documents
- I can take a particular solution aspect and clearly convey the point via one or many complimentary UML model perspectives.
The hard part is not stuffing so much into a single visual model as to render it useless. There’s no shame in requiring two (or more) different perspectives on a problem.

So there you go. Is Visio/PowerPoint/other-boxes-and-lines-tool-of-choice your first place to go when you have to diagram something for your project team or management? Or do you try and shape and morph traditional diagramming notations to fit your needs?
November 24, 2009
Validating Incoming Data Using the BizTalk Business Rules Engine
A project team recently asked me if they could use the BizTalk BRE to validate incoming data. I asked what sort of validation we were talking about, and it came down to four areas:
- Setting default values when the incoming fields were empty
- Doing look-ups based on existing data and populating related fields
- Doing concatenation of fields
- Catching any business validation errors and recording them
Before answering “of course it can”, I figured I’d quickly roll up my sleeves and prove that the BRE can do those things fairly easily.

The Setup

First, I had to get my environment set up. This started with a BizTalk XSD schema to represent the data coming into the BRE. In this scenario, the content is actually metadata about a physical file that arrives into the company.

Notice that I have a “validation errors” record at the bottom with a repeating “error” element.

Next, I have a really simple database table that can be used for lookups. Based on providing a “study” and “product” value (which comes in via the inbound XML), I want to be able to figure out which “investigator” to add to the message.

Next I need a .NET class to perform a few functions that the BRE doesn’t provide for me out of the box. First, I need to be able to concatenate two strings. Yes, it’s ridiculous that I have to write an operation to do this vs. the BRE having something built in. Deal with it.
```
public class RuleFunctions
{
   public string ConcatValues(string input1, string input2)
   {
      return input1 + input2;
   }
}
```
Next, I need a function that can add nodes to my XML message when validation errors are discovered. After adding references to the Microsoft.RulesEngine.dll and System.Xml, I wrote the following quick-and-dirty code.
```
public void AddValidationError(TypedXmlDocument activeDoc, string err)
{
   XmlDocument doc = activeDoc.Document.OwnerDocument;

   XmlNode errorRoot = doc.SelectSingleNode("//ValidationErrors");

   XmlElement newError = doc.CreateElement("Error");
   newError.InnerText = err;

   errorRoot.AppendChild(newError);
}
```
I accept a “TypedXmlDocument”, yank out the underlying XmlDocument, grab a pointer to the “ValidationErrors” record, and append a new node containing whatever error message was defined within the BRE.

After GAC-ing that helper DLL, I’m ready to roll. Before building my rules in the BRE, I wanted to establish a friendly vocabulary for my XML nodes, database columns and helper functions. So, I defined a new vocabulary and created definitions for all the “get”, “set” and “function” operations that my rules required.

Setting Default Values

The first requirement was to be able to set default values on nodes. This one’s pretty easy. I just check for either an empty string or null value, and set the default value.

Doing Lookups

The next requirement was to take values from the inbound message, perform a lookup and fill in an empty node from the message. In the rule below, I say that if the value in the XML message equals a value from the database, and another value also equals a value from the database, then set the XML node (Investigator Name) equal to another column in the database.

What this generates behind the scenes when you execute the rule is this:

So the proper T-SQL is built and executed to grab my lookup value.

Performing Concatenation

There’s no out-of-box string concatenation function in the BRE, so here I use the custom class I built earlier. I want this rule to fire at all times, so my condition is set to something that will always be true (unless math itself starts failing). I take a static file path location and then append the file name of the document coming into my company and turn that whole thing into an absolute path to the file.

Catching and Recording Business Exceptions

The final requirement was to detect business validation errors and record them. So, we can use my custom built function from earlier to create rules that look for business errors and add nodes to the message itself which outline the error details.

Remember that my custom function takes in a “TypedXmlDocument” type, so I pass the document itself into the function by dragging the root node of my schema from the XML Schemas Fact Explorer view into the operation parameter.

The Result

So what happens when I pass in a document that flexes all these conditions? I start off with an instance document that only has a few values.

So I should see some default values get set, the “investigator” node set based on database lookups, the “file path” node be populated with the concatenated value, and I should see some errors at the end because nodes like “document sub type” and “site” are empty.

To test this rule set, I’m using the built in rule tester, so I have to pass in a valid XML instance, a database connection, and a fact creator that provides the BRE with an instance of my custom .NET class.

When the rule policy is finished executing, my input XML is changed to this:

You can see that all the rules we built earlier got applied successfully. So yes, you can use the BRE to do some pretty easy-to-maintain business validation. This may not be a fit if you’re dealing with a nightly batch of 25,000 records, but if you are doing a messaging based solution, the BRE can do some nice work for you.
November 11, 2009
My Presentation from Sweden on BizTalk/SOA/Cloud is on Channel9

So my buddy Mikael informs me (actually all of us), that my presentation on “BizTalk, SOA and Leveraging the Cloud” from my visit to the Sweden User Group is finally available for viewing on Microsoft’s Channel9.

In Part 1 of the presentation, I lay the groundwork for doing SOA with BizTalk, and, try to warm up the crowd with stupid American humor. Then, in Part II I explain how to leverage the Google, Salesforce.com, Azure and Amazon clouds in a BizTalk solution. Also, either out of sympathy or because my material was improving, you may hear a few more audible chuckles. I take it where I can get it.

I had lots of fun over there, and will start openly petitioning for a return visit in 6 months or so. Consider yourself warned, Mikael.

November 3, 2009