The client was running a complex self-hosted infrastructure on Hetzner, including Kubernetes and multiple supporting technologies. The environment was unstable, difficult to operate, and caused frequent production incidents that affected application performance and user experience. We first stabilized the existing infrastructure to reduce immediate production risks, while in parallel preparing and executing a full migration to Google Cloud Platform. After the migration, we rebuilt the CI/CD process, introduced proper monitoring and alerting, implemented zero-downtime deployments, and replaced fragile self-hosted components with managed cloud services.
Challenge
The client's infrastructure was built on a fragile self-hosted stack that was only partially reliable. Kubernetes, internal services, and supporting technologies were frequently causing production incidents, slow troubleshooting, and poor user experience. The team needed to keep production running while fixing the existing environment and moving the entire platform to a more stable cloud foundation.
Objective
Reduce production instability without blocking ongoing development. At the same time, the infrastructure had to be redesigned so deployments became predictable, monitoring became transparent, and developers could work with safer and more independent environments.
What we delivered
Stabilized the existing infrastructure to reduce immediate production risks
Executed a full migration of the platform to Google Cloud Platform
Replaced unstable self-hosted components with managed cloud services such as Cloud SQL, Redis, and Memcache
Moved and standardized CI/CD with GitHub Actions
Automated deployments with Argo CD and introduced a zero-downtime deployment process
Implemented a full monitoring and alerting setup based on GCP metrics, Grafana, Prometheus, and kube-prometheus-stack
Delivered dynamic development environments that developers can create and delete independently through automated CI/CD workflows
Outcome
We stabilized the existing platform, migrated the infrastructure to GCP, moved CI/CD to GitHub Actions, automated deployments with Argo CD, and introduced zero-downtime release processes. Managed cloud services replaced unstable self-hosted components, while monitoring and alerting gave engineers real-time visibility into system health.
- Infrastructure-related production incidents reduced to a minimum
- Production became significantly more stable with predictable, zero-downtime releases
- Engineers can see the real-time state of the platform, investigate issues faster, and react to incidents before they affect users
- Developers gained automated dynamic environments they can manage independently
Technologies
GCP
Kubernetes
GitHub Actions
Argo CD
Kafka
Redis
Memcache
Grafana
Prometheus
BigQuery
ClickHouse
Cloud SQL
Hetzner
