I used to be dogmatic. I used to think that there was a “right” way and a “wrong” way of doing things. I also used to think that for the sake of the “truth” I had to shut people down when they were doing something “wrong”. It was the “right” thing to do. Retrospectively, I was probably not an easy person to work with. Now, I think it’s fairly difficult to be positively collaborative with someone who will jump at me at the hint of something unclear or inaccurate. But, there’s other types of difficult relationships. One dynamic I find quite interesting is the dynamic of a consultant and client ..read more

Recently, I have been suffering from incredibly slow pipelines and slow feedback loops. When a change to be deployed to a test environment, it would have to be approved from a pull request, then pass a validation build, merged to the master, and then deployed through a CI/CD pipeline. The problem? The whole process takes anywhere from 30-45 minutes. It feels incredibly unproductive. When any new task is added to the pipeline, it adds to the overall execution time. With the accumulation of tasks, build times can get out of hand. It’s significant because the slower the feedback loop is, then the ..read more

Based on feeling and experiencing the differences of heavy usage with Terraform and ARM, it feels that Terraform is much faster than ARM templates. So I decided to create a simple test. The Test The test created 10 app services and their app service plans. There were four cases: Terraform - When resources don’t already exist Terraform - When resources already exist ARM Template - When resources don’t already exist ARM Template - When resources already exist Here’s the source code of the tests. The results Terraform Creation (1 minute 14 seconds) Terraform Update (1 minute 15 seconds) ARM Cr ..read more

I’m not a town planner. I did go through a phase of being pretty addicted to a game called City Skylines. In this post, I’ll draw some connections in my “pretend town planner experience” with building complex multi-application infrastructures using code. In this game, you start with a huge field. In this field, you’ll have a highway with traffic already flowing. The idea is that you will build a town where there’s people coming into your city and staying. On of the first things they do is walk you through creating a road that connects to the highway. This way, newcomers can start to come into ..read more

Given the corona virus craze, a lot more people are considering working from home. I wanted to share my set-up since I spent considerable time designing my home office. I believe in spending good money in some key areas. A proper posture and work-setup will help me avoid physical pain later. A good set-up will help me stay healthier and more engaged when working. The sound My home office was pretty echoey. When I would hop on calls with co-workers, it annoyed me how I used to hear my own echo. It threw me off. The sound bounces off hard surfaces much more easily than cloth or other softer sur ..read more

One of the last sections of the 2019 State of DevOps report is how to spread the knowledge throughout an organization. Through this vast research study, they’ve asked about what approaches were used in the disseminating DevOps. They matched this data with the various organizational and SDO performance levels (low, medium, high, and elite). Turns out, commonly methods like big bang initiatives, mashups, and Centers of Excellence (CoE) are favored by low performing organizations. I was surprised about the Center of Excellence. Credits to 2019 State of DevOps Report. Page 74. Here’s the breakdown ..read more

Big Bang releases are when teams work several months on a project with the intention of pushing it live in one big event. These types of releases usually never go well. It is because there’s so much Work In Progress (WIP) that it inevitably creates an influx of unplanned work. Engineers have a trouble estimating how much it’s going to take to bring everything together. It’s almost unplannable. Emotions run high, rash decisions are made, and blame goes around. Ultimately, things usually never work out as you’d expect. Even if you happen to make it to production on a timely manner, live data and ..read more

Some time ago I wrote about how to create a YAML Azure DevOps pipeline to deploy a .NET framework v1 webjob. In this post, I will speak to a problem I had to troubleshoot. The Azure Pipelines agent timing out because it took over an hour to deploy to a continuous web job. My problem was that the continuous webjob had the binaries locked and therefore I could not upload new binaries while the webjob was running. The solution was relatively simple. Use the App Service Manage task to stop the webjob before deploying, then deploy, then start the webjob. The pipeline First, I will assume some var ..read more

Similar to a previous post, this is a small guide on how to create some glue code to ensure Flagger is installed on a Kubernetes cluster. Flagger is a delivery tool that integrates with service meshes like Linkerd to enable progressive rollouts using techniques like canaries, A/B testing, and Blue/Green deployments. It also integrates with chat apps like Slack and Teams. Pre-Requisites A YAML pipeline using the ubuntu-latest pool A pipeline that can issue kubectl commands. This post has some clues on how to connect to an AKS cluster using the Azure CLI. A service mesh installed on the cluster ..read more

GitOps allows us describe our running environment from a Git repo. We are able to state the deployments and the Helm charts that should be part of a Kubernetes cluster. As such, we can declaratively specify the version of these deployments and charts that ought to exist in the cluster. But, throughout development, it would be cumbersome to continue to increment image tags each time we want to deploy to a non-prod environment. And, we might not even want to do this for production either. Flux automated upgrades help with this. Pre-Requisites Flux installed on the cluster Helm operator installe ..read more