Atlassian Runs 300+ Clusters and Powers 2M Monthly AI Users Using OpenSearch
Customer: Atlassian
Industry: Enterprise Software and Collaboration Tools
Deployment Scale: 300+ clusters, 2,000+ nodes, 20B+ documents
Challenge: Legacy PostgreSQL-based search couldn’t scale with growing AI demands
Platform: Custom built OpenSearch platform on Kubernetes
Results:
- 99.99% availability across 300+ clusters
- 2M+ monthly AI-powered search users
- 45% reduction in developer PR cycle time
- 20B+ documents managed in largest clusters
Executive Summary
Atlassian, the company behind Jira, Confluence, and Trello, migrated critical workloads to OpenSearch. As demand for artificial intelligence (AI) and contextual search increased, OpenSearch became a foundation for new innovations such as Rovo, Atlassian’s AI solution, and RovoDev CLI, an AI developer tool.
With a Kubernetes-native infrastructure and centralized control plane, Atlassian now runs more than 300 OpenSearch clusters across over 13 global regions. The result is scalable and consistent search, platform-wide observability, and enterprise-grade compliance – achieving 99.99% availability. These capabilities support AI-driven experiences for more than 2 million users each month and empower Atlassian teams to innovate quickly, respond to customer needs, and build scalable solutions.
“OpenSearch is so flexible. It allows us to power many different scenarios… We shifted Jira’s query processing language to OpenSearch and have seen significant scalability and performance gains.”
— Fernando Garcia Valenzuela, Cloud Storage Engineering Leader, Atlassian
The Challenge
Atlassian’s legacy architecture, which relied on PostgreSQL for Jira search, could not keep up with growing scale and performance requirements. Migrating to OpenSearch provided more flexibility but introduced new challenges.
For example, Jira required read-after-write consistency, which is not native to OpenSearch. Atlassian also needed to support hundreds of thousands of tenants efficiently, but OpenSearch’s memory usage model limited the number of indexes that could exist in each cluster. Operating across multiple clouds and regions introduced additional complexity, especially around compliance with encryption key management and data residency.
Observability was another concern. The AWS OpenSearch Service exposed only cluster-level and node-level metrics, making it difficult to isolate issues or measure tenant-specific behavior.
The Solution
Atlassian addressed these challenges by building a fully managed, Kubernetes-based OpenSearch platform along with a centralized control plane. This gave engineers the flexibility to standardize operations, enforce consistency, and scale with confidence.
Migrating Jira’s query engine to OpenSearch delivered immediate performance gains. To meet Jira’s consistency requirements, the architecture was carefully designed to provide read-after-write behavior. Custom routing strategies were implemented to map tenants to shards efficiently, improving recall while minimizing latency.
Rovo, Atlassian’s AI-powered solution, was introduced to enhance search across the enterprise. It combines semantic, lexical, and structured search using OpenSearch. Powered by the Teamwork Graph, Rovo helps users find relevant information across 50 SaaS applications. It delivers AI-generated answers that respect user permissions and context.
Atlassian also developed RovoDev CLI, an AI assistant that accelerates development workflows. RovoDev CLI reduced pull request cycle times by 45 percent by streamlining code reviews and surfacing potential issues earlier in the development process.
Operating at Scale
Atlassian now supports more than 300,000 customers across over 13 regions, including government and isolated cloud environments. Its OpenSearch platform includes more than 300 clusters and 2,000 data nodes. Some of the largest clusters manage over 20 billion user-generated documents.
To manage this scale, Atlassian built a centralized control plane. It enforces security policies, manages index changes, supports customer-managed encryption keys, and ensures compliance with data residency requirements. This control plane also powers safe, consistent rollouts across all environments.
The OpenSearch infrastructure is fully orchestrated using Kubernetes. Atlassian uses open source technologies like Argo CD and Karpenter for cluster and node operations, along with Prometheus and OpenTelemetry for observability. To improve resilience, the platform takes advantage of availability zones and pre-provisioned volumes to recover quickly from failures.
These investments enabled Atlassian to achieve 99.99 percent availability while operating hundreds of clusters with stateful services.
Results
By adopting OpenSearch, Atlassian unlocked AI innovation and improved operational scale. Rovo now powers enterprise search across dozens of SaaS platforms, while RovoDev CLI enhances developer productivity through faster and higher-quality pull requests.
With a centralized control plane and standardized Kubernetes platform, Atlassian continuously delivers new platform features and maintains enterprise-grade reliability. Automation and observability improvements support safe and efficient operations across a highly complex infrastructure.
Most importantly, the platform empowers Atlassian teams to innovate quickly, respond to customer needs, and build solutions that scale.
Why It Matters
Atlassian’s journey highlights what is possible when OpenSearch is used as the foundation for scalable, AI-ready infrastructure. OpenSearch enables multi-cloud deployments, supports compliance across regions, and serves as the backbone for AI-driven products.
For organizations looking to modernize their search stack or build next-generation assistants, Atlassian’s experience shows that OpenSearch can scale to meet demanding enterprise requirements.
Watch the session:
Check out the OpenSearchCon 2025 presentation to hear directly from Atlassian engineers about how they built and operate OpenSearch at global scale.
Technical Highlights
- More than 300 OpenSearch clusters and over 2,000 data nodes in production
- Some clusters manage more than 20 billion user-generated documents
- Rovo delivers AI-powered search across 50+ SaaS applications
- Kubernetes-native orchestration with Argo CD, Karpenter, Prometheus, and OpenTelemetry
- Achieved 99.99 percent availability across stateful workloads
- RovoDev CLI reduced pull request cycle time by 45 percent
- Control plane handles encryption, data residency, compliance, and index management
