KB
Storage & Filesystems

Linux Storage & Filesystem

5 min read926 words1 code block

At a glance#

  • Purpose: Reference on Linux storage concepts: device naming, partition tables, filesystems, mounting, LVM and troubleshooting.
  • Applies to: Any Linux distribution.
  • Risk: Low as a reference - individual commands shown are destructive and marked as such.
  • Time: Reference article.

1. Storage Concepts#

Device naming in Linux#

  • Explanation:

Linux treats everything as a file, even disks. Whole disks appear as /dev/sda, /dev/sdb. Partitions are numbered (/dev/sda1). NVMe SSDs use /dev/nvme0n1 (disk) and /dev/nvme0n1p1 (partition).

👉 Important because you must know whether you’re formatting a whole disk or just a partition.

  • Demo:

``bash lsblk -f ``

Shows disks, partitions, filesystems, and mount points.

Virtual vs physical disks#

  • Explanation:

In VMs or cloud, disks are virtual devices, but Linux handles them the same as physical disks. Admins must understand this to troubleshoot when a cloud disk gets detached or resized.

  • Demo:

``bash lsblk ``

Compare physical server (sda, sdb) vs VM (often shows VDA/NVMe).

Partition table formats: MBR vs GPT#

  • Explanation:
    • MBR: Old format, supports up to 2TB disks, 4 partitions max.
    • GPT: Modern, supports huge disks, stores backup tables, nearly unlimited partitions.

👉 Use GPT for new systems unless legacy BIOS needs MBR.

  • Demo:

``bash sudo parted /dev/sdb print sudo parted /dev/sdb mklabel gpt ``

Kernel detection of new disks#

  • Explanation:

When adding new storage, Linux logs detection in the kernel ring buffer. Useful to confirm if OS can see the disk.

  • Demo:

``bash dmesg | grep sd ls /sys/block/ ``

2. Filesystem Types#

ext4#

  • Explanation: Default Linux FS. Reliable, supports large files, shrinking possible. Good general-purpose choice.
  • Demo:

``bash sudo mkfs.ext4 /dev/sdb1 ``

xfs#

  • Explanation: Enterprise FS, great performance, scales well to TB–PB. Can’t shrink.
  • Demo:

``bash sudo mkfs.xfs /dev/sdb2 ``

tmpfs#

  • Explanation: In-memory FS. Extremely fast but volatile (lost on reboot). Used for /tmp, cache.
  • Demo:

``bash sudo mount -t tmpfs -o size=100M tmpfs /mnt/tmpfs df -hT | grep tmpfs ``

Choosing FS based on workload#

  • Logs → xfs (handles many writes).
  • Databases → xfs (performance) or ext4 (shrinkable).
  • Media/archive → ext4 (simple, safe).

3. Mounting & Fstab#

Temporary vs persistent mounts#

  • Explanation:
    • mount = temporary, disappears on reboot.
    • /etc/fstab = permanent, system auto-mounts at boot.

Using UUIDs is safer than /dev/sdX since disk order can change.

  • Demo:

```bash sudo mount /dev/sdb1 /mnt/data findmnt /mnt/data

blkid /dev/sdb1 # Add to /etc/fstab: UUID=<uuid> /mnt/data ext4 defaults 0 2 sudo mount -a ```

Mount options & troubleshooting#

  • Explanation:

Options like noexec, ro, nodev improve security.

If a mount is busy (process still using it), umount may fail. Use lazy unmount (-l) or force (-f).

  • Demo:

``bash tail -f /mnt/data/file & sudo umount /mnt/data # will fail sudo umount -l /mnt/data ``

4. Disk Usage & Monitoring#

Checking usage#

  • Explanation:
    • df -h → space usage.
    • du -sh → size of directory.
    • lsblk → disk layout.
    • findmnt → see mounted filesystems.
  • Demo:

``bash df -hT du -sh /var/log/* lsblk -f findmnt ``

Inode exhaustion#

  • Explanation:

Even if free space exists, if inodes (file metadata entries) run out, you can’t create new files. Common with apps generating tons of small files.

  • Demo:

``bash mkdir /mnt/inodetest sudo mount -t tmpfs -o size=50M tmpfs /mnt/inodetest for i in $(seq 1 50000); do touch /mnt/inodetest/file$i; done df -i /mnt/inodetest touch /mnt/inodetest/newfile # will fail with "No space left on device" ``

ncdu tool#

  • Explanation:

ncdu is a curses-based interactive disk usage analyzer. Great for exploring space hogs.

  • Demo:

``bash sudo apt install ncdu -y sudo ncdu /var/log ``

5. Troubleshooting Storage#

Disk full#

  • Explanation: Services fail if no free space.
  • Demo:

``bash dd if=/dev/zero of=/mnt/data/bigfile bs=1M count=500 df -h /mnt/data ``

Deleted file still holding space#

  • Explanation: If a process has a file open, deleting it won’t free space until process ends.
  • Demo:

``bash tail -f /mnt/data/bigfile & rm /mnt/data/bigfile df -h /mnt/data lsof | grep deleted ``

Filesystem corruption#

  • Explanation: Power loss or crashes can corrupt disks. fsck repairs. Never run on mounted FS.
  • Demo:

``bash sudo umount /dev/sdb1 sudo fsck /dev/sdb1 ``

Swap usage#

  • Explanation: Swap extends RAM. If system swaps heavily, performance suffers.
  • Demo:

``bash swapon --show free -h ``

6. LVM (Logical Volume Manager)#

Why LVM?#

  • Explanation:

Lets you resize disks, combine multiple disks, take snapshots. Flexible and enterprise-grade.

LVM Workflow Demo#

bash
sudo apt install lvm2 -y
sudo pvcreate /dev/sdc
sudo vgcreate vgdata /dev/sdc
sudo lvcreate -L 500M -n lvdata vgdata
sudo mkfs.ext4 /dev/vgdata/lvdata
sudo mkdir /mnt/lvm
sudo mount /dev/vgdata/lvdata /mnt/lvm
df -hT /mnt/lvm

Extend LV#

  • Explanation: Grow disks without downtime.
  • Demo:

``bash sudo lvextend -L +200M /dev/vgdata/lvdata sudo resize2fs /dev/vgdata/lvdata df -hT /mnt/lvm ``

Shrink LV (advanced, risky)#

  • Explanation: Only ext4 supports shrink, must unmount first.
  • Demo:

``bash sudo umount /mnt/lvm sudo e2fsck -f /dev/vgdata/lvdata sudo resize2fs /dev/vgdata/lvdata 300M sudo lvreduce -L 300M /dev/vgdata/lvdata sudo mount /dev/vgdata/lvdata /mnt/lvm ``

Snapshot#

  • Explanation: Take point-in-time copies, useful before patching or risky operations.
  • Demo:

``bash sudo lvcreate -s -L 100M -n lvsnap /dev/vgdata/lvdata sudo mount /dev/vgdata/lvsnap /mnt/snap ls /mnt/snap sudo umount /mnt/snap sudo lvremove -y /dev/vgdata/lvsnap ``