How do you scale a telemetry platform when the cost of infrastructure begins to outpace the value of the data? At Sonos, the ELK stack was generating more operational burden than business value with recurring outages, permanent data loss from corrupted indexes, and on-call engineers spending up to half a sprint each month just keeping it running, making the case for migrating to AWS OpenSearch Managed Service. This wasn't just a platform swap, it was a total operational overhaul.
This talk covers the trade-offs of the transition, from decommissioning a brittle Kubernetes shipper layer to adopting a user-driven data strategy using Index State Management (ISM) and UltraWarm storage to maintain high performance while slashing operational costs by 60%.
Attendees will walk away with:
- How to design indexes, shards, and aliases before ingestion begins.
- A practical ISM rollover strategy that keeps your cluster healthy without manual intervention.
- A framework for monitoring cluster and pipeline health to empower on-call engineers with zero-dependency diagnostics.
- Hard-won lessons from migrating saved objects between Kibana and OpenSearch.