This talk shares my experience leading an SRE team, scaling from three people
running all of the company’s infrastructure to a product-centric Platform
Engineering team building tools for 200 other engineers.
I’ll explain the challenges we faced at each stage – problems that are
common across engineering orgs – and the pragmatic steps we took to solve
them, helping us get value before going all-in on an Internal Developer
Platform.
Configuring Go error reporting to work well with Sentry can be challenging. How we improved our setup by deduplicating stacktraces.
Asked for my most impactful code contribution, I discuss a small change that
had short and long-lasting impact, far beyond what I originally intended.
Go doesn't have sum types, but generics get us one step closer to a useful
polyfill. If you've ever wanted exhaustive case statements, this post is for
you.
Using a sync cutoff when building our dbt models helped fix our flaky tests,
making our CI much more reliable and exposing latent bugs.
Data should be a commodity, and accessible to everyone in a company. We've
picked a data stack for tools that make this possible- here's what we use,
how it's configured, and why we picked it.
Tie together backend and frontend using type-safe API clients, allowing the
TypeScript and Go compilers to prevent typos and developer errors.
A pattern for mocking external client libraries that keeps code simple and ensures everything down a callstack uses the same mock client.
When in a Postgres transaction, the data you read can change underneath you.
Did you realise this? Many don't, for good reasons.
Fast build times have a number of benefits, from reliability to developer
happiness. We reduces our time to deploy to <5, and it's glorious.