Company: USA TODAY (Gannett)
Industry: Media & Entertainment
Gannett | USA TODAY powers one of the world’s largest digital media networks, reaching over 200 million monthly visitors through the USA TODAY Network. Its infrastructure spans Kubernetes-based microservices, real-time publishing pipelines, and high-traffic web platforms that surge during breaking-news events like national elections. Managing observability across this dynamic, high-cardinality environment is essential not only for site reliability and performance, but for delivering trusted journalism at global scale.
Challenge
Gannett’s digital publishing operations run at extreme scale, especially during breaking-news events (e.g., a U.S. presidential election). They faced:
- Fragmented observability tooling and vendor sprawl, making monitoring, alerting and incident response inconsistent.
- Massive telemetry ingestion and cardinality growth (from many teams, clusters, legacy Prometheus instances) with growing cost and complexity.
- User/role/permissions sprawl: dozens of dev teams, legacy dashboards and alerts (“dashboard clutter”), weak traceability and ownership.
- The need to reliably scale for peak events—without disruptions, outages or surprise costs.
Solution
Gannett adopted Grafana Cloud as their unified observability platform. Key elements included:
- Consolidation: Metrics, logs, traces, alerting, on-call workflows into one Grafana Cloud-powered platform.
- Infrastructure-as-Code (IaC): Using Terraform + YAML to provision teams, groups, data-sources, roles, dashboards, users.
- Robust role & permission model: Default role “viewer”, plus custom roles (developer, IRM), managed via Active Directory groups and Terraform.
- Telemetry architecture: Prometheus in its own namespace (Kube State Metrics, node-exporter), Grafana Agent, OpenTelemetry collectors. Use of sampling and adaptive telemetry to control volume.
- Usage and cost management: Setting up metric-ingest alerts, leveraging Adaptive Metrics, usage monitoring alerts in Grafana Cloud to avoid surprise billing.
- Performance testing for high-scale events: Use of Grafana Cloud k6 alongside Grafana to simulate traffic (millions of users) and validate readiness.
“During this period, we reviewed and enabled Adaptive Metrics. Leveraging Adaptive Metrics, and the cardinality dashboards helped us reduce overall consumption, and we were even able to find a team that was inadvertently blowing up our cardinality.”
– Joseph Kregloh, Senior Manager of Cloud Engineering
Impact
Within ~3 weeks after contract signing they onboarded the company-wide Grafana Cloud deployment and user provisioning.
- Stability during peak: On “biggest day in news” (the 2024 U.S. presidential elections) the platform performed without alarms or disruptive incidents: “nothing happened. No alarms, no issues, no reports of ‘Hey this doesn’t work’.”
- Cost/control improvements: By leveraging Adaptive Metrics and sampling, they brought metric ingestion under control, avoiding quota/billing overruns and improving transparency.
- Operational consolidation: Vendors reduced (on-call, alerting, monitoring unified), dashboards and alerts cleaned up, ownership and traceability improved.
- Better team enablement: With IaC, SSO, and auto-provisioning of teams/groups, new teams could self-service observability, reducing manual overhead and silos.
“When election night came — the biggest day in news for us — everything just worked. No alarms, no outages, no drama.”
– Joseph Kregloh, Senior Manager of Cloud Engineering
Your guide
