Network-Based Read-Only Root Filesystem with High Availability
The article outlines a Linux server architecture that mounts a read‑only root filesystem over the network from a centrally managed, DRBD‑replicated control node with Heartbeat failover, providing real‑time updates, high security (even root cannot modify), and automatic migration via virtual IP.
This article describes a Linux server system with a read-only root partition that provides enhanced security, making it impossible even for the root user to modify the system. This read-only root filesystem (or customized operating system) is actually a client-server system where the OS image is mounted from a central control node over the network.
Core Single Point: The central control node is a single point of failure that stores one or more customized operating system images and provides various services. If this node fails, the entire cluster would collapse. To avoid this single point of failure, the system uses DRBD (Distributed Replicated Block Device) for real-time data synchronization and Heartbeat from the Linux-HA project for automatic failover. The system achieves second-level automatic migration with virtual IP support.
Key Features:
Centralized management - only one operating system needs to be maintained
Real-time updates - any modifications take effect immediately
High security - even root user cannot compromise the system
Exception handling - allows specific files to exist on client servers
Use Cases:
MMORPG servers - providing only computation and database services
High security environments - root user cannot perform malicious actions
Other scenarios requiring enhanced system integrity
Setup Steps:
Deploy DRBD and Heartbeat
Recompile Linux kernel
Customize the operating system
Configure TFTP service
Configure DHCP service
Configure NFS service
PXE network boot
Key Technical Points:
I. Kernel Compilation: Module parameters only work after kernel boots, so NFS must be compiled directly into the kernel (not as a module) to support Root file system on NFS. The system uses Ramdisk to place /var/tmp in memory for proper operation of system services.
II. System Boot Process:
Client boots via network card (PXE) and automatically discovers DHCP service
Downloads customized kernel via DHCP and TFTP
Obtains IP address and starts the operating system
Moves /var and /tmp to memory (tmpfs)
Logs into the operating system
Verifies NFS root mounting
Even attempting "rm -rf /" fails due to read-only nature
Baidu Tech Salon
Baidu Tech Salon, organized by Baidu's Technology Management Department, is a monthly offline event that shares cutting‑edge tech trends from Baidu and the industry, providing a free platform for mid‑to‑senior engineers to exchange ideas.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.