Glitches in the Matrix, or Taming Agent Chaos
summary
Fleet manages osquery agents with concrete mitigations: a process watchdog that blocks expensive queries for 24h, Linux cgroups as a hard backstop, per-query profiling and Fleet UI metrics for real-world cost, and agent-side buffering/backpressure so endpoints retry until server ack. Test infrastructure uses simulated agents (tens to hundreds of thousands per VM) to exercise agent check-ins and MySQL/Redis failure modes, plus an automated fleet control debug archive command to collect traces; these practices reduce incidents and shorten debugging cycles.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 1:05What is osquery
- 4:15Prioritizing risks
- 6:41Mitigating agent risks
- 9:32Fleet server reliability
- 14:45Self-managed challenges
- 16:49Consistent deployments
- 19:13Automation and testing
- 22:43Debugging and response
- 24:29Key takeaways
related talks
31mObservability in the Matrix: Choosing the Right AI Agent To Escape the Chaos
32mWhat Agents Do When You’re Not Looking
17mFighting AI with AI
19mLive Q&A
29mThe osquery Agent. Performance Tuning @Scale
01h 20mMind the Gap (In your Agent Observability)
47mValidation and Observability of AI Agents
40mChaos Engineering at Scale