Ironic Operator Feedback
summary
Operators report Ironic runs production workloads for telecoms and handles benchmarking, burn‑in, health checks, retirement and handing out nodes but users resent the extra abstraction when nodes go into clean failed and must open tickets. Proposed fixes include read-only API access or scoped actions, using the owner field to map nodes to projects for visibility, and exposing unified limits from Placement (quota per resource class, hopeful for Wallaby). Work in review: snapshot spec, kickstart/anaconda driver; RAID demand remains low.
AI-written summary · about
chapters (9)
- 0:00Introduction and setup
- 1:51Operator feedback session begins
- 6:07User acceptance challenges with Ironic
- 17:41Unified limits and resource class quotas
- 24:49Snapshot and RAID support discussions
- 30:48Exploring alternative backends
- 34:04Ironic as multi-tenant cloud interface
- 42:40RAID layering and future plans
- 43:22Closing remarks and Q&A
related talks
28mControllers: Lambda Functions for Extending your Infrastructure
32mEvent-Driven Infrastructure: Autonomous Operation and Unreliable Networks
29mProactive Ops: Event Driven IT Operations
43mAutomated Testing & Board Farming
24mBOSS: Building Operating System Services
24mCan't Touch This!: When Things go Wrong With an Appliance
15mThe Ops in Serverless
34mOpen Source in Safety Critical Applications: the Next Frontier