Beyond the Ephemeral: Mastering Serverless Metrics at Scale With Shopify
Talkat Observability Day 2025 · on Tuesday, 2025-04-01· at 16:30· Level 1 | Hall Entrance S10 | Room C·
summary
Shopify builds a serverless metrics pipeline that runs OpenTelemetry collectors as sidecars, sending OTLP over the public internet into regional ingestion clusters and then through a metric router into Prometheus‑compatible storage. It converts exponential histograms to DD Sketches, handles 100 GB/s logs, 12 million spans/s and ~350 million stat samples/s at peak, and uses reduced collector flush intervals, 10s Cloud Run shutdown handling, and label‑dropping cardinality control to prevent data loss and runaway cost.
AI-written summary · about
chapters (17)
- 0:00Serverless metrics at Shopify
- 1:45Global observability challenges
- 3:34Building Observe platform
- 5:22Kubernetes at Shopify
- 6:23Motivation for serverless
- 8:10Serverless observability challenges
- 9:14Serverless metrics instrumentation
- 10:19StatsD and OpenTelemetry
- 12:26OpenTelemetry sidecar
- 13:52Metrics pipeline and traces
- 14:38Metric router and cardinality
- 16:04Sidecar config for traces
- 17:49Day two operations
- 18:53Developer experience
- 19:34Logging and security next
- 20:38Pipeline optimizations
- 21:21Q&A
related talks
03mUsing OpenTelemetry and Prometheus to monitor serverless
34mObservability in Serverless Application
23mShopify's journey from conventional to native histograms
34mFleeting Metrics: Monitoring Short-lived or Serverless Jobs with Prometheus
21mMonitoring Serverless Workloads with OpenTelemetry and Prometheus
56mFull-stack observability and application monitoring with
26mScaling OpenTelemetry for Modern Workloads: From Mobile To Ma...
41mResilient Observability at the Retail Edge: A Lightweight, Scalable, and...