Operating OpenSearch at Scale: Lessons From Managing 10,000+ Clusters Across...
summary
Run a K8s-based platform that manages 11,000 clusters and ~60,000 persistent volumes, uses custom Kubernetes operators plus per-namespace maintainers to automate lifecycle tasks, and implements disk-based autoscaling, zone-aware routing and ingestion with Fluentd and OpenTelemetry collectors. Faced large-scale update and API-throttling failures, introduced progressive, throttled rollouts and caching for IDP metadata, and reports clear operational improvements in provisioning, scaling and recovery.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:51SAP BTP observability needs
- 8:16Telemetry pipeline
- 13:5811k OpenSearch clusters scale
- 24:44Kubernetes operator roles
- 34:00Infrastructure challenges
- 40:10Operator maintainer separation
- 48:29Identity provider alerting optimizations
- 50:34Future outlook
related talks
39mOperating OpenSearch at Scale With Kubernetes Operators: Lessons Learned and Community Insights
40mOperating Opensearch - The Kubernetes Way
14mKeynote: OpenSearch on Kubernetes at Atlassian
21mNoOps for OpenSearch: The Future of Self-Healing Search Infras...
23mCreating the OpenSearch Kubernetes Operator
28mOpenShift for the Enterprise or What Could Possibly Go Wrong
06mSponsored Keynote: What I Learned Running 10,000+ Kubernetes Clusters
38mMigrating from Self-Managed ElasticSearch to OpenSearch