The situation
The platform is a public source of medicines information, relied on by healthcare professionals and the public. It had to be available all the time, and it had to be possible to change it safely.
When we started, development was done by another company. Deployments were rare, and they often caused outages.
Built to survive failure
We designed a fault-tolerant cloud architecture on AWS, spread across multiple availability zones, so that a server failing, or even a whole zone having problems, didn't take the platform down.
Making releases safe
We put development and QA environments in place, and every deployment had to pass through them and be tested before it reached production. We insisted that QA was an exact replica of production at all times: after seeing deployments cause outages, it was the only way to be sure a release that worked in testing would work live.
We built the deployment automation ourselves, and used exactly the same process for every environment, so production deployments were never the first time a process ran. Once this was in place, deployment problems were caught before they reached production.
Taking over development
In 2022 we also took over development of the application itself (a Symfony API with an Angular front end), with our agency partner acting as product owner and QA. With one team responsible for both the code and the platform, we could release far more often: up to twice a day at times, and about once a week on average, with zero downtime.
We also regularly replaced the entire infrastructure underneath the platform without an outage or maintenance window beyond a few seconds.
The result
99.999% uptime over the platform's final full year with us, about five minutes of downtime in a year, despite frequent deployments. The engagement ended in 2026, when the industry body moved to a provider that already offered the same service as part of a wider platform.