Disk volumes troubleshooting

Partition table corrupted (disk shows wrong size or no partitions)

Backup first: sfdisk -d /dev/sda > sda-backup.txt (do this BEFORE any changes). Use fdisk -l /dev/sda to check visibility. Try parted /dev/sda print to see if GPT backup header is intact. For GPT, restore backup header: gdisk /dev/sda, then r (recovery) and then b (use backup header). For MBR, testdisk /dev/sda can search for lost partitions. Notice that the package testdisk might need to be installed to perform this last task.

Partition/Volume UUID conflict after cloning (Filesystem has duplicate UUID or wrong fs type, bad option, bad superblock)

To safely use the clone alongside the original volume, the filesystem UUID on the cloned volume must be changed to a new unique value. With the cloned volume attached to a working server, identify the clone's device name using lsblk. Assuming the device name is /dev/nvme2n1p1, mount to clean the journal mount -o nouuid /dev/nvme2n1p1 /mnt/clone, then umount the volume with umount /mnt/clone and regenerate the UUID based on the filesystem type: For XFS do xfs_admin -U generate /dev/nvme2n1p1 and for ext2/3/4 do tune2fs -U random /dev/nvme2n1p1. Verify the new UUID is unique with blkid | grep nvme2. Now /etc/fstab can be updated with the clone volume and the volume with the new UUID can be mounted normally with mount -a.

Filesystem full (No space left on device)

Run df -hT to find the full mount. Use du -xhd1 <mount> to locate large directories. Check for deleted-but-open files with lsof +L1 — space is not reclaimed until the process closes the file.

Inodes exhausted (space available, writes fail)

Check df -i. Millions of tiny files (mail queues, caches, sessions) can fill inodes first. Find dense directories with find /path -xdev -type f | cut -d/ -f1-4 | sort | uniq -c | sort -n or remove/rotate the offending data.

Mount missing after reboot

Inspect /etc/fstab for typos, wrong UUID, or missing nofail on non-critical mounts. Run mount -a to test. Check boot logs: journalctl -b | grep -i mount.

mount: wrong fs type, bad option, bad superblock

Verify filesystem type with blkid or lsblk -f. A mismatched fstab type (ext4 vs xfs) or corrupted superblock causes this. Run fsck on an unmounted filesystem if needed.

target is busy on umount

A process still has files open under the mount. Find it with lsof +f -- /mount/point or fuser -vm /mount/point. Stop the service or use a lazy unmount only when you understand the risk: umount -l.

Filesystem remounted read-only

Often I/O errors or a full disk triggered a protective remount. Check dmesg | tail for disk errors. Free space, fix hardware issues, run fsck, then mount -o remount,rw /.

LVM: extended disk but filesystem unchanged

Storage expansion may involve up to four layers, all of which must be checked and potentially resized.:

Confirm each step with lsblk, pvs, lvs, and df -h.

Device name changed (/dev/sdb became /dev/sdc)

Kernel enumeration order can shift when disks are added. Prefer UUIDs in fstab:

Debugging workflow

1. Map the storage stack

2. Find what's consuming space

3. Validate fstab before reboot

Emergency space cleanup (safe patterns)

Avoid deleting files from /proc, /sys (they are virtual filesystems — nothing there takes disk space and you can't meaningfully "delete" files from them anyway.), or random paths under /var/lib without knowing what owns them.