06. Operating System
Chapter focus:
🧭 Reading Map
- Prerequisites: 03. Cluster Architecture — hardware topology and network planning determine how the OS is installed and configured
- Role: opening of Part 2, L2 — turning bare metal into a "homogeneous, reproducible, rollback-able" OS environment: PXE autoinstall, diskless boot, Golden Image pipelines, system networking (NAT) and boot troubleshooting
- Order: provisioning pipeline & image pipeline (big picture) first, then PXE troubleshooting and system networking
- Next: 07. Scheduling & Cloud — once machines are up, who decides which GPU goes to whom
Articles
Cluster Delivery Automation: From PXE Provisioning + Ansible Config to Slurm-Ready
Reproducible flow and scripts dissecting cluster delivery: node acceptance after PXE, Ansible batch config (network/storage/drivers), Slurm readiness checks, with a delivery checklist and engineering takeaways.
阅读文章 →
Incident Review: How LD_LIBRARY_PATH Broke dracut and XFS Mount
A GPU/HPC server failed to boot with the error: mount: unknown filesystem type ''xfs''. The root cause was not disk failure, but environment variable po
阅读文章 →
Diskless Linux Boot via PXE + NFS (Ubuntu / Rocky / Red Hat)
In data centers, cluster environments, educational labs, or large-scale node deployments, Diskless Boot is a common and efficient way to start systems
阅读文章 →
Ubuntu 22.04/20.04 Automated Provisioning: Cloud-init & Subiquity
A complete guide to automated bare-metal provisioning for Ubuntu Server using Subiquity, Cloud-init (Autoinstall), and dual-mode UEFI/Legacy boot.
阅读文章 →
PXE Troubleshooting Handbook for Large-Scale Provisioning
six classic failures: symptom, root cause, fix, and a decision tree.
阅读文章 →
Large-Scale Provisioning Pipeline: Golden Image, Version Management & Canary Rollout
reproducible builds, versioning, canary strategy.
阅读文章 →
Linux Dual-NIC NAT Gateway Guide
At the OS level, configure a dual-NIC Linux node as a NAT gateway so internal nodes share public access; covers iptables, Firewalld and nftables.
Read →