OpenSearch at Atlassian Scale
summary
Atlassian runs OpenSearch across 300+ clusters, 2,000+ data nodes, and some clusters with 20 billion documents to serve 300,000 customers and 2 million monthly AI users. The platform team built a control plane on Kubernetes using Argo CD, Carpenter, Prometheus and OpenTelemetry, migrated Jira query processing from PostgreSQL to OpenSearch and ships Robo in six months. Multi-tenancy forces single-index routing, shard-level visibility gaps in AWS managed service, and automation plus tailored observability are essential to keep latency low and availability high.
AI-written summary · about
chapters (14)
- 0:00OpenSearch at Atlassian
- 1:04Jira’s migration challenges
- 2:28Teamwork Graph explained
- 3:32OpenSearch infrastructure numbers
- 4:14Multi-tenancy challenges
- 5:16Control plane for OpenSearch fleets
- 6:19Cultural shift to centralized infra
- 7:02Balancing AI growth and tech debt
- 8:04Custom OpenSearch on Kubernetes
- 8:47Kubernetes stateful service complexities
- 9:49Observability at scale
- 10:31OpenSearch community support
- 11:15Future of OpenSearch at Atlassian
- 12:38Q&A
related talks
13mSecuring the Data Frontier: Real-World Strategies With OpenSearch
16mA Thrilling Journey into Data-Driven Security Automation
41mSecuring OpenSearch: Hints for avoiding common hassles
38mOpen-source Dashboarding Tools. Which One Suits You?
38mEnabling Secure, Fine-Grained Resource Sharing for Team Collaboration in OpenSearch
55mFirewall Reversed: How Open Source Became Asia's Best Defense
14mOpenSearch Data Separation With JWT Authentication for Multi-tenant Applications
25mThe Art of the Possible with OpenSearch Anomaly Detection and UEBA.