From Single-Server Shops to a Highly Available E-Commerce Platform: How ayedo Restructured Nordlicht Commerce

E-commerce agencies often grow faster than their infrastructure.

Nordlicht Commerce specializes in Shopware projects for medium-sized retail companies and D2C brands. Twelve customer shops in operation, six proprietary plugins in the Shopware Store, a growing project business – strong in expertise, technically committed.

What worked for a long time became a risk with an increasing number of customers and higher SLA requirements.

Initial Situation: Twelve Shops, Five Hosts, Zero Standardization

Over the years, a typical agency setup had developed.

Each shop ran on its own managed hosting package with different web hosts. PHP versions varied. MariaDB configurations differed. Redis was present in some shops, absent in others. Elasticsearch versions were not uniform.

Each shop was technically unique.

For smaller projects, this was manageable. With increasing demands, it became an operational burden.

If a plugin worked for Customer A, it didn’t mean it would run stably for Customer B. Debugging consumed a disproportionate amount of time – not due to faulty code, but due to inconsistent environments.

Even more critical was the issue of availability.

Each shop ran on a single server. No failover, no horizontal scaling. During maintenance or traffic peaks – such as on Black Friday or campaign launches – performance issues or outages were inevitable.

A customer planned a TV campaign and asked if the shop could handle ten times the traffic. The honest answer was: probably not.

In parallel, there was a lack of a clean development environment for the plugin teams. Locally, testing was done with SQLite, without Redis, without Elasticsearch, without queue systems. Problems only appeared after deployment – at the customer.

Staging environments existed only for a few major customers. The rest went directly into production.

Deployments were done via SSH and manual scripts. Rollbacks were manual work. In case of a failed update, the responsible developer was stuck working late into the night.

The turning point came when two major customers simultaneously demanded SLAs with 99.9% availability. Single-server hosting could not deliver that.

The question was no longer “How do we optimize?” but: How do we fundamentally professionalize?

The Strategic Approach: A Central E-Commerce Platform Instead of Twelve Individual Installations

Our goal was not to improve the existing servers. Our goal was to build a platform.

A standardized, highly available Kubernetes-based e-commerce infrastructure where all shops are consistently operated – isolated, scalable, and automated.

The Solution: Shopware as a Scalable Workload on ayedo Managed Kubernetes

We built a central managed Kubernetes platform for Nordlicht, where each customer shop runs as an independent deployment.

The key difference: All shops share the platform – but not their resources.

Isolated Shops with Individual Scaling

Each shop runs in a dedicated namespace with its own resource limits. Horizontal Pod Autoscaling ensures that additional pods are automatically started during traffic peaks.

Whether a regular weekday or Black Friday – the platform dynamically adapts.

Last Black Friday, the largest shop automatically scaled to eight times its normal capacity – without manual intervention.

Highly Available Data and Backend Services

Databases run as highly available clusters with automatic failover and point-in-time recovery. Redis handles session management and caching. OpenSearch or alternatively Typesense ensures performant product searches. RabbitMQ decouples time-critical shop operations from background jobs like email sending or indexing.

The result: No single server as a single point of failure anymore.

Self-Hosted Additional Services Instead of SaaS Dependency

A significant lever was the integration of additional services directly on the platform.

PDF generation via Gotenberg. Geocoding and routing with Nominatim and OSRM. Server-side tracking via a self-hosted Google Tag Manager container. Mail testing with Mailhog. Secure data exchange via SFTPGo. AI-driven text generation via a local LLM (Ollama).

This not only eliminated external SaaS costs for Nordlicht but also data protection risks from US services.

All services run on European infrastructure – fully controlled.

GitOps Instead of SSH

ArgoCD handles the deployment of all shops. Shop updates and plugin releases are triggered by Git commits and automatically rolled out. Health checks verify integrity. In case of problems, an automatic rollback occurs.

No more manual SSH deployments. No more late-night emergency rollbacks.

Production-Identical Development Environments

For the plugin team, this was a quantum leap.

Each plugin branch can automatically start a complete shop instance with MariaDB, Redis, OpenSearch, RabbitMQ, and all additional services – identical to the production environment.

Errors are detected before they occur at the customer.

Development no longer happens blindly, but under realistic conditions.

The Key Lever: Standardization

The greatest effect came from unification.

Instead of twelve individual server setups with five hosts, there is now a unified platform. Configurations are identical. Differences only exist in shop-specific parameters like domain, theme, or plugin configuration.

A new customer shop is provisioned from a template and is production-ready within an hour.

What used to take a week of lead time is now a standardized process.

Result: Stability, Scalability, and New Business Opportunities

Nordlicht now achieves contractually guaranteed 99.9% availability.

Plugin bugs after deployment have decreased by over 80% because tests take place in realistic environments.

Deployments take minutes instead of hours – with automatic rollback.

Monthly costs for external SaaS services have decreased by around 60%.

New customer projects can be accepted faster because infrastructure is no longer a bottleneck.

And perhaps most importantly: Nordlicht can now confidently accompany campaigns with traffic peaks – instead of cautiously slowing them down.

Why This Approach Works

E-commerce agencies often underestimate how much infrastructure influences the quality of their projects.

Operating twelve shops individually scales linearly with complexity. Operating a standardized platform scales with processes.

Kubernetes provides isolation and scaling. GitOps ensures reproducibility. Self-hosted additional services create sovereignty. Managed operation relieves the team.

This is how an agency becomes a platform operator.

Call to Action

If you operate multiple Shopware or e-commerce projects and struggle with inconsistent environments, manual deployments, or lack of high availability, this is not an individual problem – it’s a platform issue.

With the ayedo Managed Kubernetes platform, we build you a standardized, highly available e-commerce infrastructure where shops, plugins, and additional services can be operated consistently and sovereignly – without US SaaS dependency and without late-night SSH rescue operations.

Let’s jointly assess how your current hosting architecture looks – and how it can become a scalable platform that enables SLAs instead of preventing them.