How do We Keep Metrics for a Long Time in VictoriaMetrics
summary
Comparison shows VictoriaMetrics storing long‑term metrics with far lower resource cost: a 12‑hour test on a 2‑CPU, 7.7 GB RAM, 25 GB disk bench ingests 344 billion points and VM keeps ~9.7 GB vs Prometheus ~49 GB while using noticeably less CPU. Uses vmagent push mode to avoid open ports and gaps, applies downsampling to shrink retention (examples: 5s → 10s → 30s → 1m → 5m → 1h) and deduplication to collapse high‑availability duplicates; single‑node works up to ~1M metrics/sec, cluster for scale.
AI-written summary · about
chapters (10)
- 0:00Start
- 4:05Push Module
- 5:45Push Model
- 12:05Downsampling
- 17:16Data Storage
- 22:57Risks to the Community
- 27:18The Abuse of Open Source
- 32:49How Developers Are Going To Impact the Future of Postgres
- 51:00What Was Your Favorite Release of Postgres
- 51:53What's Your Favorite Food
related talks
40mScale Your Monitoring Solution With the VictoriaMetrics Ecosystem
36mVictoriaMetrics: scaling to 100 million metrics per second (EN)
30mHow We Scaled Metrics Monitoring for 10,000 Applications With Victoria...
33mAggregating metrics in-flight: challenges and opportunities
34mSpecifics of Data Analysis in Time Series Databases
35mWorking with filesystem in Time Series database
50mHow To Improve PostgreSQL Monitoring and Observability With PMM
28mThe Sixth Sense: Finding “Dead” Metrics and Silent Alerts in Your Monitoring