Atlassian Migrated Metrics Pipeline to OpenTelemetry
The company overhauled its massive telemetry stack to handle billions of data points per minute with increased efficiency.
Updated on Sept. 29, 2026 in Data Centers
Live Poll
Do you believe transitioning legacy software systems to modern open-source standards generally improves long-term reliability?
Atlassian has successfully migrated its internal metrics infrastructure from gostatsd to the OpenTelemetry Collector. The upgraded system now processes 4.8 billion data points per minute across 100,000 hosts while maintaining a 99.95% service level objective.
Why it matters
By adopting OpenTelemetry, Atlassian eliminated the limitations of its previous UDP-only pipeline, enabling future integration of logs and traces. The transition allows the company to leverage community-developed features rather than maintaining custom legacy code.
The metrics platform currently manages 4.8 billion incoming data points per minute across 14 regions, storing 220 million points per minute. The new architecture achieves a 96% volume reduction and cuts CPU usage by 50% compared to the legacy implementation.
The players
Atlassian
A global enterprise software company known for products including Jira, Confluence, and Trello, currently focused on optimizing internal cloud infrastructure.
The details
The platform is divided into collection, ingest, aggregation, and forwarding stages to allow for modular upgrades. To maintain continuity, the team kept the existing StatsD-over-UDP (User Datagram Protocol, a connectionless communication method) interface, preventing the need for immediate application-wide re-instrumentation. The new ingest system uses a load-balancing exporter to hash data by stream ID, while a custom aggregation processor handles delta metrics—data reflecting change over time—before final storage.
Timeline
September 29, 2026: Publication of the metrics migration report.
The Tech Race
The transition reflects a broader shift among large-scale cloud operators to standardize observability stacks using the OpenTelemetry Collector. Atlassian is effectively moving away from bespoke sidecar deployments to align with an industry-standard framework capable of unifying metrics, logs, and traces.
Users of Atlassian services will see increased reliability as the platform sustains a 99.95% service level objective despite a massive processing load. For developers, the move signals an impending transition toward using native OpenTelemetry SDKs rather than legacy StatsD or vendor clients.
The takeaway
The move validates OpenTelemetry as a viable standard for massive-scale internal metrics processing. Watch for the next phase, where the company plans to move individual application instrumentation away from legacy clients and toward unified OpenTelemetry SDKs.
Further reading
For broader trends in infrastructure management, see our latest analysis of Data Centers.
Source note: This article includes information reported by InfoQ.
Live Poll
Do you believe transitioning legacy software systems to modern open-source standards generally improves long-term reliability?
