Case study · Engineering services · Digital platform operator
Observability for a 200,000-user platform on GKE
One view of health across dev, test and production, load-tested at 200,000 concurrent users.
200Kconcurrent users in load testing
The problem
What the client needed
Dev, test and production environments running Cassandra, Kafka and Cloud SQL had no single view of health.
The result
What runs now
Production pushes its metrics out, so nothing in the non-production cluster can reach into production.
GKEGrafanaPrometheusArgoCDCassandraKafkaCloud SQL
What we built
The system, step by step
1
A two-cluster topology: a non-production cluster hosting Grafana and a collector for dev and test, and a production cluster with a push-only collector
2
Prometheus for metrics and ArgoCD for GitOps delivery
3
Load testing at 200,000 concurrent users
Have a problem you want measured?
Tell us what you want to count, check or protect. An engineer will reply, not a sales script.