Five years of Observability at Canonical
After five years of leading Observability at Canonical, that journey is coming to an end.
Simon Aronsson
Recent content on Simon Aronsson
Latest posts
"Why is everything so slow?" - a case study
A Juju charm performance case study tracing slow NRPE target reconciliation to repeated downstream relation databag updates, and showing how batching cuts hook-tool invocations dramatically.
What I'm doing now.
Building A personal operating system called Loop, helping me be more present, focus on high-leverage work, without dropping anything unintentionally. Reading Mostly technical and organizational material around engineering leadership, cloud-native systems, observability, and practical AI. I’m trying to bias toward things that change how I work, not just things that are interesting.
The end of "Just ask Sarah"
Documentation used to back up organizational memory. With agents, it becomes execution context.
Boring code is an organizational tell
Why clever code is an organizational output, not an engineering one — and how agents turn a slow-moving problem into a fast one.
What happens if you give an AI agent a home instead of a session?
My design goals for building an autonomous AI agent, rather than limiting it to interactive prompting session-by-session.
COS Lite Machine Sizer
A sizing tool for COS Lite deployments. It used to live here but got lost in one of my many blog migrations. Now it’s back!
Building a Dry-Run Mode for the OpenTelemetry Collector
Signal Studio explores a deficit in the OpenTelemetry ecosystem: how to assess the impact of changes to your config.yaml before rolling out in production.
How I built this page
This site has needed a facelift for years. Not because the technology was outdated, but because every previous version of this blog eventually died. Quietly.
Observability is becoming mission critical, but who watches the watchmen?
The last couple of years, there has been quite a lot of development lowering the barrier of entry for observability. There are now quite a few, reasonably mature options out there that lets you set up a good monitoring stack either through a few clicks or by a few one-liners in the terminal.
Error Economics - How to avoid breaking the budget
At SLOConf 2021 I talked about how we may use error budgets to add pass/fail criterias to reliability tests we run as part of our CI pipelines.
On shaming and blaming
There are severe consequences of allowing shaming and blaming in your engineering culture. In this essay, I’ll suggest a few small things you can do instead.
About
What I work on I’m an engineering leader from eu-north-1 working at Canonical, where I focus on making observability accessible and valuable for teams of any size. My background is in open-source, developer communities, and building teams that ship. I spend most of my time on the boring end of the problem: how to keep systems understandable as they grow, how to make on-call something a team…