Refactoring everything everywhere all at once: Safety A/B testing in single-tenant SaaS
At Octopus Deploy, we rebuilt the core engine of our product - without the usual chaos of a big-bang release, upset customers and a one-way door, or a cartesian explosion of feature flags. Instead, we built a safety analysis engine that decides in real time which version to run based on what a customer’s doing and what the new engine supports, to enable our developers to ship continuously. How did we pull that off - and make sure no customer was broken twice? Come find out.
Who is this talk for?
SREs/Ops folks, Software developers from junior to principal.
What will attendees take away?
Some things to think about when starting a major rearchitecture, especially if they’re in a single tenant cloud platform, specifically:
When to start thinking about how they want to roll it out, to avoid painting themselves into a big-bang-release shaped corner.
Some ideas around designing safety nets to reduce customer impact from changes.
Bringing some careful thinking around where and when to use feature flags for large projects, and the downsides of not pre-planning how your flags will intersect.
The difficulties of a single tenant Cloud platform, or any large set of diverse environments and how to think about managing all the variation and complexity upfront, instead of being bitten by it later.
Alix is a lead engineer at Octopus Deploy. She has primarily worked at NZ product tech companies including Pushpay, Auror, Sharesies and as a co-founder of Multitudes.
She has previously been involved in the NZ developer community as a co-organiser for GDG and C# Meetups in Auckland, co-running an iteration of DevFest, organising and participating in Women and Non-binary People in Tech initiatives, and speaking at conferences.