--- title: Atlassian Rebuilds Metrics Pipeline Around OpenTelemetry... description: Atlassian has detailed its migration of an internal metrics platform from gostatsd to OpenTelemetry, covering roughly 100,000 hosts across 14 regions while... canonical: https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Atlassian Rebuilds Metrics Pipeline Around OpenTelemetry at Massive Scale | daily.dev og:description: Atlassian has detailed its migration of an internal metrics platform from gostatsd to OpenTelemetry, covering roughly 100,000 hosts across 14 regions while... og:url: https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques og:image: https://api.daily.dev/og/posts/Iayb6queS.png og:image:alt: Atlassian Rebuilds Metrics Pipeline Around OpenTelemetry at Massive Scale og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Atlassian Rebuilds Metrics Pipeline Around OpenTelemetry at Massive Scale **[InfoQ](https://daily.dev/sources/infoq)** · 5 min read · 0 upvotes · 0 comments ## Summary Atlassian has detailed its migration of an internal metrics platform from gostatsd to OpenTelemetry, covering roughly 100,000 hosts across 14 regions while maintaining a 99.95% SLO. Instead of forcing services to switch clients, the platform team preserved the existing StatsD-over-UDP interface and rebuilt the pipeline behind it using purpose-built OpenTelemetry Collector distributions for collection, ingest, aggregation, and forwarding. Key results include a roughly 30% cut in sidecar cost, a switch to stream-ID-based load balancing to fix hot-shard problems, a custom open-sourced aggregation processor that reduces 4.8 billion data points per minute down to about 220 million (a 96% reduction) while using half the CPU, and a stateless metrics-gateway for forwarding to destinations like SignalFx and S3. The piece also contrasts Atlassian's approach with Airbnb's dual-emission migration to vmagent and Skyscanner's library-based standardization on OpenTelemetry, and includes commentary from GREE's monitoring lead praising the design while questioning the scalability of the stateful aggregation tier. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much did Atlassian reduce metrics data volume by moving aggregation to OpenTelemetry Collector? Atlassian cut data volume by roughly 96%, going from about 4.8 billion data points received per minute down to about 220 million stored, using a custom open-sourced aggregation processor built because upstream OpenTelemetry components did not aggregate delta metrics the way the company required. The revised aggregation tier also uses about half as much CPU for the same traffic. _Teams sizing their own metrics pipeline can compare notes on aggregation trade-offs like this on daily.dev._ ### How did Atlassian migrate from gostatsd to OpenTelemetry without disrupting production alerting? Atlassian kept its existing StatsD-over-UDP interface unchanged and rebuilt the pipeline behind it in four stages: collection, ingest, aggregation, and forwarding, using purpose-built OpenTelemetry Collector distributions for each. This turned what could have been an organization-wide application migration into a project owned solely by the platform team, while preserving the 99.95% SLO used by production alerts. _Engineers planning an observability migration under strict SLOs can follow approaches like this on daily.dev._ ### What load balancing approach did Atlassian use to fix hot shards in its metrics ingest tier? Atlassian replaced its old nomad proxy, which hashed by service and environment onto single shards, with the OpenTelemetry contrib load-balancing exporter that hashes by stream ID instead. This keeps individual time series together while spreading one large service's traffic across the whole pool, producing more even CPU distribution, fewer hot shards, and better off-peak scaling. _Anyone debugging uneven shard load in a stateful pipeline can dig into solutions like this on daily.dev._ ## Similar posts on daily.dev - [OpenTelemetry everywhere: Migrating a metrics platform at scale](https://daily.dev/posts/opentelemetry-everywhere-migrating-a-metrics-platform-at-scale-xaaglqltp) · CNCF · 0 upvotes · 0 comments - [Airbnb Migrates High-Volume Metrics Pipeline to OpenTelemetry](https://daily.dev/posts/airbnb-migrates-high-volume-metrics-pipeline-to-opentelemetry-ytqavyfqp) · InfoQ · 1 upvotes · 0 comments - [Elastic’s new metrics capabilities will dramatically improve uptime for public sector IT](https://daily.dev/posts/elastic-s-new-metrics-capabilities-will-dramatically-improve-uptime-for-public-sector-it-1tv2tpxd0) · elastic · 0 upvotes · 0 comments - [Inside Adobe's OpenTelemetry pipeline: simplicity at scale](https://daily.dev/posts/inside-adobe-s-opentelemetry-pipeline-simplicity-at-scale-xbmirj76w) · OpenTelemetry · 7 upvotes · 0 comments - [How to build a cost-effective observability platform with OpenTelemetry](https://daily.dev/posts/how-to-build-a-cost-effective-observability-platform-with-opentelemetry-z8pbichl4) · CNCF · 8 upvotes · 0 comments --- Tags: [#observability](https://daily.dev/tags/observability), [#distributed-systems](https://daily.dev/tags/distributed-systems), [#opentelemetry](https://daily.dev/tags/opentelemetry), [#atlassian](https://daily.dev/tags/atlassian) [View this post on daily.dev](https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Atlassian Rebuilds Metrics Pipeline Around OpenTelemetry at Massive Scale","url":"https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques"},"datePublished":"2026-09-29T09:06:48.666Z","dateModified":"2026-09-29T09:07:13.907Z","description":"Atlassian has detailed its migration of an internal metrics platform from gostatsd to OpenTelemetry, covering roughly 100,000 hosts across 14 regions while...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/54dcfdec617c02586c9760846edd04ff?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/54dcfdec617c02586c9760846edd04ff?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"InfoQ","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"InfoQ","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/afc3bced3e1e4b188dd9127017a60e0c","url":"https://daily.dev/sources/infoq"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"observability,distributed-systems,opentelemetry,atlassian","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"InfoQ","item":"https://daily.dev/sources/infoq"},{"@type":"ListItem","position":3,"name":"Atlassian Rebuilds Metrics Pipeline Around OpenTelemetry at Massive Scale"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/atlassian-rebuilds-metrics-pipeline-around-opentelemetry-at-massive-scale-iayb6ques#faq","mainEntity":[{"@type":"Question","name":"How much did Atlassian reduce metrics data volume by moving aggregation to OpenTelemetry Collector?","acceptedAnswer":{"@type":"Answer","text":"Atlassian cut data volume by roughly 96%, going from about 4.8 billion data points received per minute down to about 220 million stored, using a custom open-sourced aggregation processor built because upstream OpenTelemetry components did not aggregate delta metrics the way the company required. The revised aggregation tier also uses about half as much CPU for the same traffic. Teams sizing their own metrics pipeline can compare notes on aggregation trade-offs like this on daily.dev."}},{"@type":"Question","name":"How did Atlassian migrate from gostatsd to OpenTelemetry without disrupting production alerting?","acceptedAnswer":{"@type":"Answer","text":"Atlassian kept its existing StatsD-over-UDP interface unchanged and rebuilt the pipeline behind it in four stages: collection, ingest, aggregation, and forwarding, using purpose-built OpenTelemetry Collector distributions for each. This turned what could have been an organization-wide application migration into a project owned solely by the platform team, while preserving the 99.95% SLO used by production alerts. Engineers planning an observability migration under strict SLOs can follow approaches like this on daily.dev."}},{"@type":"Question","name":"What load balancing approach did Atlassian use to fix hot shards in its metrics ingest tier?","acceptedAnswer":{"@type":"Answer","text":"Atlassian replaced its old nomad proxy, which hashed by service and environment onto single shards, with the OpenTelemetry contrib load-balancing exporter that hashes by stream ID instead. This keeps individual time series together while spreading one large service's traffic across the whole pool, producing more even CPU distribution, fewer hot shards, and better off-peak scaling. Anyone debugging uneven shard load in a stateful pipeline can dig into solutions like this on daily.dev."}}]} ```