Running the Observability As a Service For Your Teams With Thanos
Talkat KubeCon + CloudNativeCon North America 2022 · on Wednesday, 2022-10-26· at 16:30· 252 AB, Huntington Place: 1 Washington Blvd·
summary
Thanos moves Prometheus to a multi-cluster SaaS model by adding concrete features: bucket-prefix multitenancy (single S3 bucket with per-tenant prefixes), remote_write and active-series limit enforcement, and Katama consistent hashing for stable series distribution. Query stack changes include pushdown, query sharding and a new parallel PromQL engine; on a 10M-series test dataset some queries fall from ~11s to ~3s and others from ~6s to ~2s, demonstrating real, deployable speed and isolation improvements.
AI-written summary · about
chapters (8)
- 0:00Start
- 4:26Thanos: Distributed Prometheus
- 13:08Bucket Prefix
- 14:03Why Rate Limiting
- 15:03Ingestion Limits (Per Tenant!)
- 21:57Query Pushdown
- 22:59Query Sharding
- 25:32Based on Volcano Paper
related talks
39mAbsorbing Thanos Infinite Powers for Multi-Cluster Telemetry
25mMulti-Cluster Observability: What you need to survive with a fleet
15mScaling Observability Stack for AIOPS
22mMonitoring the World: Scaling Thanos in Dynamic Prometheus Environments
26mPrometheus That Scales
27mThanos: Highly Available, Pluggable, Long Term Metric Storage for Everyone!
23mConnecting Thanos to the Outer Rim via Query API
10mUpstream-First, High Scale Prometheus Ecosystem