AGENT SKILL

vmexit-analysis

暂无简介

License 许可
Apache-2.0
最近更新
2026-11-15
Stars
★ 0
Forks
3

03 SKILL.md 正文

/vmexit-analysis

Analyze KVM vmexit statistics to identify virtualization overhead and configuration issues.

Methodology

  • Read methodology/vmexit-analysis.md for vmexit interpretation guide (HLT vs IO_INSTRUCTION, count vs time impact)

Data Source

VM exit stats from KVM debugfs:

cat /sys/kernel/debug/kvm/*/vcpu*/vmexit > vmexit_stats.txt

What to analyze

  • HLT exits: High time % indicates guest is idle — expected for idle VMs, investigate if workload should be CPU-bound
  • IO_INSTRUCTION exits: High count on Windows VMs indicates useplatformclock issue (force Windows to use PIT)
  • EPT_VIOLATION: Memory pressure or hugepage issues
  • EXTERNAL_INTERRUPT: Interrupt storms or poor IRQ affinity
  • PAUSE_INSTRUCTION: Lock contention, possible CPU overcommit

Key principle

Count ≠ Time impact. Example:

  • IO_INSTRUCTION: 68% of exits, but only 13% of time
  • HLT: 32% of exits, but 87% of time ← this is the critical exit

Always report both count % and time % when analyzing vmexits.

Steps

Automated analysis script: Not yet implemented — manual analysis workflow:

  • Collect vmexit stats from KVM host
  • Calculate count % and time % for each exit type
  • Sort by time % (not count %) to identify critical exits
  • Cross-reference with workload type (CPU-bound vs I/O-bound vs idle)
  • Correlate with domstats metrics (see /io-analysis)
  • Provide diagnosis and fix recommendations

Common fixes

  • IO_INSTRUCTION high: Inside Windows VM run bcdedit /deletevalue useplatformclock then reboot
  • EPT_VIOLATION high: Disable memory balloon (autoattachMemBalloon: false), enable hugepages
  • PAUSE high: Reduce CPU overcommit, set requests = limits for guaranteed QoS

Integration

  • Windows VM clock config: see methodology/vm-tuning-guide.md Steps 9-10
  • CPU allocation: see methodology/vm-tuning-guide.md Steps 7-8
  • Hugepages: see methodology/vm-tuning-guide.md Step 6

04 同仓库更多 Skills