Hello sweet internet! At this point, you should know that I am far to be a lovely happy code. I like to share what I learn and to have a chat about what you are doing. That’s as it is! Feel free to follow my wheeze on twitter/gianarb.
If you don’t know what I am gonna speak about, I can tell you this is another way to enjoy coding!
Recently a friend of mine that organizes the CloudConf in Turin, Italy asked me if I was able to deliver a workshop. Let me say THE workshop, 8 hours of chatting with exercises and questions. I did something like that years ago about AngularJS but hey, this sounds like a challenge, and I love challenges! So I took it. If you read my recent posts you know I have a passion nowadays:
The topic was clear, I have called it “Application instrumentation”. Lovely!
I am driven by passion and purpose. My passion for troubleshooting and the purpose of figuring out what the f happen in production. I was ready to work on it!
This article is about how I prepared the workshop and I hope it can help somebody to avoid the same mistake and also to use some of the material I developed.
I made everything in open source. There are two new repositories on my GitHub one with fake e-commerce I made using 4 different programming languages:
You can find the code on github.com/gianarb/shopmany.
I decided to develop a minimum version of the application in order to have it reusable for another purpose. It can be used to build a use case for Kubernetes deployment for example or a CI lesson.
master contains the minimum set of features that I need to have an
application that has some sense. But for example, the services are without logs,
metrics and tracing because they will be added as exercises from the attendees.
If you check out the workshop you will be able to see in the history a commit for every exercise and applications.
The lessons are available on github.com/gianarb/workshop-observability, every directory is a lesson. The readme contains a couple of information about what where we are, why we should care and one or more exercise to do in practice in order to familiarize with the concepts.
The lessons I developed for the purpose of the CloudConf workshop are:
I have also reported an idea of a possible timeline (the one I used at the CloudConf):
09.00 Registration and presentation 09.30 - 13.00 Theory
13.00 - 14.00 Launch 14.00 - 17.00 Let’s make our hands dirty 17.30 - 18.00 Recap, questions and so on
I like to prepare slides, posts, and workshop because I learn a lot along the way about concepts that I usually develop during a long set and frustrating attempts. Or reading a lot of blog posts, books, code. Writing about it helps me to put together what I learned developing easy to understand materials.
This workshop was not a special case. It is not clear for me that even if there is a lot going on with OpenCensus, OpenTracing, and other instrumentation libraries there is still room for improvement.
Instrumenting an application is not anymore just a matter of adding
around the execution of the code. But it is the way we have to write an
application capable of behind debugged and that speaks with the outside in an
The course has two different sections: theory and practice.
The theory went well. I do not have a lot to say about it and for me, it is where I am most comfortable with because it looks like a long talk.
The practical part was for sure a bit too long and I didn’t have time to walk all the people over it but the fact that there are all the solutions, the purpose written down helped them to feel less lonely and everyone can follow the resolution is it can do the exercise in practice.
This usually happens because of different skills set or for trouble configuring the environment.
git helped me a lot, every commit has a diff that I used to explain the
solution of the lessons. People that were not confidently writing the solution in a
particular language had to just
cherry-pick the commit in the language they
The practical part was designed to be a collaboration between people. IMHO it helps to feel less “at school” but more as a team that it is something we should feel more comfortable with at work.
I think it worked but not that well. People were supporting and helping each
other. But I probably need to cut the lessons in a different way. I think I will
influxdb lessons injecting the learning process of only what
matters for the course along the other lessons. Next time and I will develop a
new lesson about how to parse the logs and push them to InfluxDB for example.
(let me know if you would like to help me!)
I asked them to do a survey before the end of the course in order to help me get their feeling. There is a lot to do and some of their feedback is part of this article. But in general, I am happy because I have all the material in order and this for me was just a first iteration. I hope to make it better, to grant more feedback from the open source and to run it again! So let me know if you would like to have me on board!
As I said instrumentation is hard and I still hoping to get an easier solution across languages. I tried OpenCensus but I didn’t manage to have it running at I was in the rush so I used Jaeger.
I will develop something about structured logging as I said for sure.
I hope to get a lesson from some as a service provider like HoneyComb for example.
The youngest person in the room was a student in high school! Wow!