A client story of migrating from self-hosted Elasticsearch to a managed OpenSearch
My conference talk with our client on the stage of OpenSearchCon’26 in Prague
UPD: case study based on the talk: https://opensearch.org/case-studies/zero-downtime-modernization-norwegian-labor-and-welfare-administrations-path-to-open-source-observability-with-opensearch/
This month I had a pleasure of sharing the stage with customer Nav in the sunny Prague at the OpenSearchCon.
Hans Kristian Flaaten (Nav), my co-presenter, has laid out the story of how Nav evolved as a software organization from consultancy-driven to product-driven. And how they become open-source and API first.
The recording:
We discussed a few lessons we have learnt together at Nav and Aiven’s side while migrating from Elasticsearch to OpenSearch.
Practical takeaways for anyone planning a similar migration:
1. Do not start with the search engine. Start with user workflows. NAV needed full-text search over logs. Loki did not satisfy that central use case.
2. Avoid historical migration when possible. Dual-write plus retention-window logic is much safer than bulk reindexing terabytes of historical logs.
3. Test the whole ecosystem, not just OpenSearch. Agents, OpenTelemetry, traces, dashboards, RBAC, alerting, and log formats matter.
4. Use migration as a cleanup event. NAV used the move to improve logging hygiene and reduce inappropriate sensitive-data logging.
5. Managed services are not magic. They reduce operational burden, but only if the provider can actually handle your scale and integration constraints.
6. Open source strategy is not just ideology. For NAV, licensing, public funding, GitHub-based openness, and operational control were part of the architecture decision.
Btw, if you would like to learn OpenSearch in practice and build a prototype, you can do so with Free Tier: https://aiven.io/free-opensearch
