Discover OpenStack's nerve with oslo.metrics: Have a robust private cloud on a large scale

on YouTube ↗26m

" summary

oslo.metrics extracts internal metrics from oslo libraries and sends them in Prometheus exposition format to Grafana so operators can see concrete RPC and DB behavior across a 2000+ hypervisor cloud with ~50,000 VMs and 20,000 bare metal hosts. It spots RPC anomalies such as resolve_block_device_name timeouts that caused duplicated volumes and quantifies scheduling latency: select_destination: 10→20s, 100→35s, 500→140s (120s timeout). A hypervisor emulator for testing scale to ~10,000 nodes is under development.

AI-written summary · about

" chapters (14)

1. 0:00Introduction and agenda
2. 1:09Line's infrastructure overview
3. 3:59Problems faced in scaling OpenStack
4. 5:23Scalability and reliability focus
5. 6:46What is oslo.metrics
6. 8:55oslo.metrics architecture
7. 14:15Usage in production environment
8. 14:57Metrics visualization with Grafana
9. 16:25Metric exchange behind operations
10. 18:09Troubleshooting with oslo.metrics
11. 19:57Trend monitoring for prevention
12. 21:02Current and future integrations
13. 22:49Hypervisor emulation project
14. 25:15Q&A and summary