Disk Design

Previous Next

Objectives

  • Master the differences between MBR and GPT partitioning.
  • Understand the Linux Filesystem Hierarchy Standard (FHS).
  • Learn to configure mount points and /etc/fstab.

1. Partitioning Theory

Understanding how data is organized on a disk is fundamental to server administration and cloud infrastructure.

MBR vs. GPT

  • MBR (Master Boot Record):
  • Legacy Standard: Introduced in the early 1980s.
  • Capacity Limit: Uses 32-bit values for addressing, limiting disk size to 2TB.
  • Partition Limit: Supports only 4 primary partitions. To exceed this, one must be configured as an "extended partition" containing multiple "logical partitions".

  • Single Point of Failure: The partition table is stored only at the beginning of the disk.

  • GPT (GUID Partition Table):

  • Modern Standard: Part of the UEFI specification.
  • Massive Capacity: Uses 64-bit addressing, supporting disks up to 9.4 Zettabytes.
  • Unlimited Partitions: Theoretically unlimited, though most OS implementations (like Linux) default to 128.
  • Redundancy & Integrity: Stores a backup copy of the partition table at the end of the disk and uses Cyclic Redundancy Checks (CRC) to detect corruption.
  • Protective MBR: GPT includes a "protective MBR" at the very beginning (LBA 0). This is a dummy MBR partition table that covers the entire disk, making legacy tools see the disk as a single, unknown partition rather than an empty disk. This prevents older tools from accidentally overwriting GPT partitions.

Inspection Tools

To list and verify partitions on a system, use these essential commands:

# List all block devices in a tree format with filesystem info
lsblk -f

# List partition tables for all disks (requires sudo)
sudo fdisk -l

# Identify a specific disk's partition scheme (e.g., /dev/sda)
sudo parted /dev/sda print

2. Linux Filesystems

A filesystem defines how data is stored and retrieved. Selecting the right one is critical for performance and reliability.

Key Filesystem Types

  • Ext4 (Fourth Extended Filesystem):
  • The most widely used filesystem in the Linux world.
  • Highly stable, supports volumes up to 1EB and files up to 16TB.
  • XFS:
  • A high-performance 64-bit journaling filesystem.
  • Excellent at handling large files and high I/O throughput; default for RHEL and CentOS.
  • Btrfs (B-Tree FS):
  • Copy-on-Write (CoW): Instead of overwriting data in place, Btrfs writes new data to a different block and then updates the metadata.
  • Snapshots: This CoW nature allows for near-instantaneous, space-efficient snapshots. If a system update fails, you can roll back the entire filesystem to a previous state instantly.
  • Swap:
  • Dedicated disk space used as "virtual memory" when physical RAM is exhausted.

The Importance of Journaling

Most modern filesystems (Ext4, XFS) use Journaling to ensure data integrity.

  1. The Problem: In a non-journaling filesystem, if the system crashes while writing a file, the filesystem structure (metadata) might be left in an inconsistent state, requiring a slow fsck (filesystem check) on the next boot.
  2. The Solution: A journal is a dedicated area (a log). Before making changes to the actual data blocks, the filesystem writes the intended changes to the journal.
  3. Recovery: If a crash occurs, the system simply checks the journal. It either completes the pending operations (replay) or ignores incomplete ones, bringing the filesystem back to a consistent state in seconds rather than hours.

3. The Filesystem Hierarchy Standard (FHS)

The FHS defines the directory structure and directory contents in Linux distributions. It ensures that software can predict the location of installed files and libraries.

Key Directories and Their Purpose

  • /etc (System Configuration):
  • Contains all of the system-wide configuration files.
  • Examples: /etc/passwd (user database), /etc/ssh/sshd_config (SSH server config), /etc/nginx/nginx.conf (Web server config).
  • /var (Variable Data):
  • Files to which the system writes data during the course of its operation.
  • Examples: /var/log/syslog (system logs), /var/lib/mysql (database storage), /var/spool/mail (mail queues).
  • /home (User Data):
  • Personal directories for regular users.
  • Examples: /home/alice/Documents, /home/bob/.bashrc.
  • /root (Root Home):
  • The home directory for the root user (separate from /home to ensure availability if /home is on a different, unmounted partition).
  • /boot (Boot Loader):
  • Static files required to boot the system.
  • Examples: vmlinuz (the Linux kernel), initrd.img (initial RAM disk).
  • /bin & /sbin (Essential Binaries):
  • /bin: Command-line binaries essential for all users (e.g., ls, cp, bash).
  • /sbin: System administration binaries essential for the root user (e.g., fdisk, reboot, iptables).
  • /usr (User Binaries):
  • The largest share of data on a system; contains user-facing programs and libraries.
  • Examples: /usr/bin/python3, /usr/share/doc.
  • /dev (Device Files):
  • Hardware representation files.
  • Examples: /dev/sda (first SATA disk), /dev/null (the "black hole").

4. Mounting & Persistence

In Linux, storage devices are not automatically available; they must be "mounted" to a directory in the FHS.

The /etc/fstab File

The File System Table (/etc/fstab) determines how disk partitions, remote storage, or other file systems are integrated into the system at boot.

Anatomy of an fstab entry

# <file system>  <mount point>  <type>  <options>  <dump>  <pass>
UUID=a1b2c3d4...  /var/log  ext4  defaults,noatime  0  2
(1)               (2)       (3)   (4)               (5)(6)
  1. <file system>: The device identifier. While /dev/sda1 works, UUIDs (Universally Unique Identifiers) are preferred (see below).
  2. <mount point>: The directory where the filesystem will be accessible.
  3. <type>: The filesystem type (e.g., ext4, xfs, nfs, swap).
  4. <options>: Mount parameters. defaults includes rw, suid, dev, exec, auto, nouser, and async. noatime is often used on SSDs or cloud instances to reduce writes by not updating "last access" times.
  5. <dump>: Used by the dump utility (legacy). Usually set to 0.
  6. <pass>: Used by fsck to determine the order of filesystem checks at boot. 1 for the root filesystem, 2 for others, 0 to disable.

The Importance of UUIDs

In cloud and server environments, Device Names are Unpredictable.

  • If you add a new disk to a virtual machine, what was /dev/sdb might become /dev/sdc.
  • If a hardware controller initializes disks in a different order, your system might fail to boot if it relies on /dev/sdX names in fstab.

UUIDs solve this by providing a unique string tied to the filesystem itself, regardless of which physical port or controller it's connected to.

How to find a UUID

# Display the UUID and filesystem type for all block devices
sudo blkid

Resources

Ecosystem Intro Boot Process