Hardware Onboarding and Burn in with Ironic in the CERN
summary
CERN automates hardware onboarding with Ironic: nodes auto-register via IPA, send introspection to S3/Ceph and populate OpenDCIM; cleaning becomes burn-in via Ironic Xena using stress-ng and FIO, with fluentd streaming logs to Elastic Search/Kibana; network burn-in uses dynamic pairing backed by Zookeeper and scales to 150 nodes in parallel in tests; deployment topology places ~500 nodes per Ironic controller and manages about 9,000 nodes, moves from IPMI to Redfish, and exposes Terraform-driven provisioning and retirement tagging.
AI-written summary · about
chapters (9)
- 0:00Introduction to CERN IT
- 1:27Hardware lifecycle management
- 2:48Ironic architecture overview
- 8:45Batch computing and Ironic
- 9:55Ironic cleaning workflow
- 11:39Burn-in and provisioning
- 23:28Growth and retirement
- 24:38Conclusion
- 25:08Q&A session
related talks
24mImage-Based Bare-Metal Provisioning With Metal3.io, Cluster API and…
50mBare Metal DevOps
29mTowards Sustainable Multi-Cluster Management in Real-World Academic Infrastructure
13mAutomate an Observable Runtime with HashiCorp and Ansible
36mOSFCI - Extensible Open-source CI for Firmware with Real Hardware Execution
36mContinuous Deployment of IoT Sensors
28mHow to Deploy CF and Other Bosh Releases on Bare-Metal Environment
24mNetworking track kick-off