← All Posts
DevOpsCI/CDAWSReliability

Zero Downtime in 90 Days: A FinTech Deployment Transformation

Sean Lobjoit··2 min read

I helped a FinTech client reduce their 4-hour maintenance window to zero downtime deployments in 3 months.

The Problem with Scheduled Maintenance

It's not uncommon for many companies in this sector to have scheduled downtime, usually on a Sunday, where every release means having engineers on call, customers seeing maintenance pages and the risk of further revenue loss if the rollout doesn't go to plan.

For a 24/7 financial platform, this isn't acceptable long term. The reasons for this are nearly always the same:

  1. Deployments are tightly coupled, everything has to ship at once.
  2. There's no proper rollback strategy in place. The only strategy is usually to re-deploy the old version, manually.
  3. Database updates are coupled to application deployments.
  4. Deployments happen less frequently due to loss of confidence which in turn makes each deployment riskier.

What We Changed

In reviewing their architecture, I introduced the following:

  1. A blue-green deployment pipeline (in AWS) that shifted traffic in under 60 seconds.
  2. I decoupled database changes from application deployments with backward-compatible schema changes (expand/contract).
  3. Feature flags allowed code to ship independent of releasing new features.
  4. Automated deployment tests that run before traffic cut-over.

The Results

After 3 months we had achieved zero downtime releases. That's not to say we didn't have deployment failures. We did. We had achieved zero downtime deployments, 100% of the time.

Deployment frequency went from once per week on a Sunday to multiple times daily. Rollback time was reduced from 4-hour full re-deployments to 45 seconds.

Infrequent Deployments Don't Reduce Risk

If there is one takeaway from the above, it's this:

Infrequent deployments don't reduce risk. They increase it. The longer you wait between releases, the more changes are piling up and as a result the harder it is to isolate what broke when things do go wrong (and they will).

Zero-downtime deployments are not only an infrastructure win, they're a culture shift that lets your teams move faster without fear.

Dealing with painful release cycles? Let's fix that.