Collecting WCF Telemetry with Application Insights

I got pulled back in to diagnosing performance issues on an old friend, and it lead me into some dead code with a chance for resuscitation.

Trouble with the WCF

One of our applications has a set of WCF Services that serve the client. We had not instrumented them manually for performance metrics, although we found that System.ServiceModel is well decorated for trace, so adding a trace listener gives us a bevy of information. However, it’s all in *.svclog files (using an XML Trace Listener), so there is still a lot to do in order to find out what’s wrong. A colleague asked if Application Insights could work. And that suggested started me down a path.

I found the lab

I reached out to some contacts at Microsoft, and they pointed me to the Application Insights SDK Labs. In that lab is a library that instruments WCF services and applications. The problem: it hasn’t been updated in about 5 years.

I figured, our application is based on technology about that old, I suppose I could try it. So I followed the instructions, and I started getting telemetry in my Application Insights instance! However, I did notice a few things:

1. The library as published relies on an old Application Insights version (2.5.0, I believe). The Application Insights libraries are up to 2.21, which means we may not see everything we can.
2. The library is based on .Net Framework 4.5, not 4.6.2, which is what we are using.

So I did what any sensible developer does…. fork it!

I’ll do it myself!

I forked the repository, created a branch, and got to work upgrading. Since this is just for work (right now), I did not bother to setup CI/CD yet, and took some liberties in assuming we would be running .Net Framework 4.6.2 for a while.

I had to wade through the repository setup a bit: There is a lot of configuration around Nuget and versioning, and, frankly, I wasn’t looking to figure that out right now. However, I did manage to get new library built, including updating the references to Microsoft.Application insights.

Again, in the interest of time and test, I manually pushed the package I built to our internal Nuget feed. I’m going to get this pushed to a development environment so that I can get some better telemetry than just the few calls that get made in my local environment.

Next Steps?

If this were an active project, I would probably have made a little more effort to do things the way they were originally structured and “play in the sandbox.” However, with no contribution rules or guidelines and no visible build pipeline, I am on my own with this one.

If this turns out to be reasonably useful, I will probably take a look at the rest of the projects in that repository. I may also tap my contacts at Microsoft for some potential “next steps” with this one: while I do not relish the thought of owning a public archive, perhaps I could at least get a copy of their build pipeline so that I can replicate it on my end. Maybe a GitHub Actions pipeline with a push to Nuget?? Who knows.