Programming the Performance Co-Pilot toolkit
summary
Claims that Performance Co-Pilot can be extended with Python to export real metric names and metadata into the PCP namespace and that a small Python pmda parses the kernel command output (dmsetup status) to produce dm-cache metrics. The pmda registers metrics with semantics, the fetch callback returns live values, and existing tools like pmcd and pm_logger plus pmchart immediately plot or archive them. Remote hosts and historical archives work and the prototype is small and functional.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:29Outline
- 4:41Overview
- 6:12Architecture
- 9:26Metrics
- 13:23Device Mapper Cache target
- 15:45Collector extensions in Python
- 22:46Monitor extensions in Python
- 26:36Resources
related talks
28mLightweight Observability with Performance Co-Pilot and Grafana
36mIntroducing a New Performance Monitoring Daemon
46mPowerShell as an MCP Tooling Host: Using MCPServerPS to Supercharge AI-Assisted...
47mBuilding a Foundation for Performance
01h 21mBeyond the Prompt: Leveraging MCP and PowerShell for Custom LLM Actions
25mLet Me Be Your OpenTelemetry Champ
47mDive into BPF performance tools using python
25mDPDK PMD Live Upgrade