Thirty services onto Kubernetes before the biggest sale of the year
How we would help an online retailer whose traffic jumps several times over during the festive sale move its services to Kubernetes, without risking the sale itself.
This is a composite example built from common enterprise requirements. It is not a specific client engagement, and the figures are design targets for the scenario, not measured results.
Over-provisioned all year, nervous all sale
The retailer ran about thirty services on virtual machines sized for the festive peak, so most of that capacity sat idle for eleven months. Releases needed a weekend and a war room. Last year, a slow recommendation service backed up checkout during the busiest hour.
What we would do
- Fit check: many services, weekly releases, big spikes and a team that already used containers. Kubernetes fits.
- Managed Kubernetes in the existing public cloud, with GitOps delivery from day one.
- Move low-risk services first (catalogue, search), then checkout and payments last.
- Autoscaling on request rate for customer-facing services, with tested upper limits.
- Timeouts and circuit breakers so a slow non-critical service cannot hold up checkout. See the API simulator.
- Load test at twice the expected peak four weeks before the sale, and freeze changes in sale week.
Targets and how they are checked
| Measure | Target | Checked by |
|---|---|---|
| Idle capacity outside the sale | Much lower | Monthly cloud bill and utilisation |
| Release effort | Routine weekday releases | Deployment frequency and rollback count |
| Checkout during peak | Unaffected by non-critical services | Load test results and sale-day monitoring |
What we would flag
- The deadline is fixed. If checkout is not ready in time, it stays on virtual machines for this sale.
- Autoscaling needs warm-up. Pre-scale before known peaks; do not rely on reacting in the first minute.
- The database is the real limit. Scaling application pods does not help if the database cannot keep up.
Planning a Kubernetes platform, or fixing one?
Tell us how many applications and teams you have, how you deploy today and where it should run. We will come back with an honest view: whether Kubernetes fits, what a lean platform looks like, and what it takes to run.