Shore Scale
Work / weather.com cache operator
94%
cache hit ratio

A custom Kubernetes operator that keeps weather.com's cache at 94%

The Weather Channel (weather.com)weather.comMedia & data
Results
Cache operations
500+ lines of scripts
→ 1
declarative Go operator
Origin load
baseline
→ 10x
less load on origin
Cache hit ratio
—
→ 94%
across both regions
Config changes
manual
→ GitOps
via ArgoCD
The challenge

Weather traffic is all spikes. A storm warning can multiply requests in minutes.

CDN caching across two IBM Cloud regions was managed by hand-written scripts that were fragile, slow to change and risky at weather.com's scale.

Architecture
100%
Switch layers on and off above, or hover a line in “What we built” to light up its layer.
What we built

Scripts out. One operator in.

  • 01 · OPERATOR

    Custom Varnish operator in Go: a leader-elected controller that watches a VarnishCache resource, generates VCL, pushes it with hot reloads, health-checks and rolls out updates safely.

  • 02 · K8S CONFIG

    Kubernetes-native config: a CRD for replicas, VCL rules, backends, TTLs and purge policies, with RBAC.

  • 03 · CACHE TIER

    Cache tier: a Varnish StatefulSet with disruption budgets and autoscaling on miss rate.

  • 04 · MULTI-REGION

    Multi-region failover: active US-South and standby EU-DE clusters behind anycast DNS and Istio mTLS.

  • 05 · CI/CD · OBSERVE

    CI/CD and observability: GitHub Actions, Helm, ArgoCD, kind end-to-end tests, Prometheus, Grafana and Jaeger.

Stack
GoKubernetes operatorCRDVarnishIstioNGINX IngressIBM CloudHelmArgoCDGitHub ActionsPrometheusGrafanaJaeger

Expecting a storm of traffic?

Talk to an engineer →
Next case
Mobile advertising · AdTech
Real-time bidding engine
11x
200ms → 18ms