Cortex: How to Run a Rock Solid Multi-Tenant Prometheus

Cortex is a CNCF open-source project providing horizontally scalable, highly available, multi-tenant, long-term storage for Prometheus. This talk delves into Cortex's architecture, resilience strategies, and features that ensure continuous metric flow. Key topics include hash-ring and replication factor for fault tolerance, zone-aware replication for outages, tenant and instance limits for cost and usage control, and shuffle sharding to minimize outage impact. The presentation also covers recent releases, including support for vertical sharding, open telemetry, ARM images, and an experimental new Prometheus query engine. The speakers emphasize the importance of measuring and maintaining reliability through SLOs and user feedback, highlighting Cortex's design for Kubernetes and its integration with Prometheus and Thanos.