Configuring Prometheus for High Performance
summary
Prometheus tuning centers on local-storage flags: set memory chunk limits so chunks in RAM don’t exceed about one third of host RAM (practical rule divide RAM by 6 rule), set max chunks waiting for persistence to about half that, and watch the exported urgency metric — hitting rush mode at 0.8 urgency forces aggressive writes and fsync tradeoffs. Expect checkpoints to grow linearly (minutes to tens of minutes on 10s of GB), and reduce Go heap growth with GOGC=40 to lower steady-state memory.
AI-written summary · about
chapters (5)
- 0:00Start
- 6:45What Determines Where the Steady-State Is
- 7:46Local Memory Chunks
- 12:35What Is Rush Mode
- 24:48What Is Checkpointing
related talks
35mPushing the Limits of Prometheus at Etsy
42mPrometheus Intro and Deep Dive
06mLearnings from Running Prometheus on Google Cloud Scale
25mUnbreakable Prometheus: Leveraging Limits for Consistent Monitoring
03mPrometheus Project Updates
36mDoing Things Prometheus Can’t Do with Prometheus
37mPractical Capacity Planning using Prometheus
16mUnlock High-Frequency Deployments without Blowing Up Prometheus