Zabbix: 100k NVPS on a Single Server
summary
Zabbix monitoring platform is examined with focus on its limits and how to scale the analytics storage and optimize the Zabbix server and UI. The talk explains an architecture and practical optimizations used to run Zabbix at over 40k NVPS (net values per second) and summarizes operational experience and key conclusions.
AI-written summary · about
chapters (27)
- 0:00Introduction
- 1:11What is Zabbix
- 2:14Monitoring challenges at Intersvyaz
- 4:24Switching to ClickHouse for metrics
- 6:34Optimizing database performance
- 6:55Slow data collection issues
- 7:58Improving availability checks
- 9:03Asynchronous polling with nmap
- 10:05Synchronous polling limitations
- 13:21Diagnosing semaphore bottlenecks
- 14:25Zabbix server data flow
- 15:30Config cache and queue competition
- 16:36Preprocessor manager optimizations
- 17:17Scaling preprocessor manager
- 18:01Metrics per second record
- 19:27Final optimizations and trade-offs
- 20:10Testing the new system
- 20:56Fixing slow widget loading
- 21:43Summary
- 24:31Q&A
- 24:52Support concerns
- 26:41Implementation details
- 27:49Replacing fping with nmap
- 29:19Zabbix as alerting tool
- 30:32Resource allocation
- 31:59Adapting for proxies
- 33:44Closing remarks
related talks
58mZabbix: Recipes for High-Performance Monitoring
46mZabbix and Millions of Metrics: Best Practices for Large-Scale Monitoring
53mZabbix 3.2 — Monitoring on a Whole New Level
47mMonitoring at Scale: Intuitive Dashboard Design
37mMonitoring as code: a GitOps approach with Zabbix
30mZabbix meets AI–The power of opensource monitoring
32mMonitoring OVH: 300K servers, 27 DCs one unified metrics platform
41mMonitoring OVH: 300k Servers, 27 DCs, a Metrics Platform