NTFS is the standard file system for Windows drives. Its heart is the Master File Table, or MFT: a database with one record per file holding its name, timestamps, permissions, and a map of where its data lives. Very small files fit entirely inside their MFT record. A journal records metadata changes so the volume can recover from a crash.
When a file is deleted, NTFS marks its MFT record as unused and frees its clusters in an allocation bitmap. The record and the data often survive until they're reused, which is why NTFS deletions on hard drives are frequently recoverable.
Corruption usually hits the MFT or the boot sector. NTFS keeps a mirror of the first MFT records and a backup boot sector at the end of the volume, which recovery tools use to restore access.
Intermediate · ntfs · windows · mft · file system
APFS has been the default file system on Macs, iPhones, and iPads since 2017. A disk holds a container, and the container holds several volumes that share its free space, such as the read-only System volume and the Data volume with your files. APFS is copy-on-write: changes go to new blocks instead of overwriting old ones, which also makes snapshots cheap.
Copy-on-write sometimes helps recovery, because older copies of metadata can remain on disk. But on SSDs, deleted blocks are trimmed quickly, and APFS volumes are often encrypted, so deleted-file recovery from internal Mac SSDs is usually very limited.
Time Machine and local snapshots are the best protection. On external hard drives formatted as APFS, a damaged container can often be read through older checkpoint copies of its metadata.
Advanced · apfs · macos · snapshots · encryption
HFS+ was the Mac's file system for many years before APFS, and it still turns up on older Macs, external drives, and Time Machine disks created before APFS took over. It stores the directory tree in a catalog file organized as a B-tree, with a separate file for extra fragment information and a journal to protect against crashes.
When a file is deleted, its catalog entry is removed and the tree is reorganized, so names and folder structure are harder to rebuild than on NTFS. Data can still be found by carving or by reading leftover catalog nodes.
Corruption often shows up as an invalid node structure error. Repeated repair attempts can rebuild the catalog in ways that drop entries, so on important drives, image first.
Intermediate · hfs+ · mac os extended · journaling · catalog file
ext4 is the default file system on many Linux systems and some NAS devices. Files are described by inodes, which store attributes and a tree of extents pointing to the data blocks. A journal protects metadata, and backup copies of the superblock are spread across the disk.
When ext4 deletes a file, it clears the extent information in the inode, so the link between name and data is broken right away. Deleted-file recovery often relies on old copies of metadata in the journal, or on carving files by their content.
Corruption is usually repaired with tools that use the backup superblocks to restore access. Running repairs on a failing drive, or on a RAID that was reassembled incorrectly, can make the damage permanent.
Advanced · ext4 · linux · inodes · journal
XFS is designed for large files and big volumes. It splits a disk into allocation groups that each manage their own space and inodes, so it can work in parallel. It's the default on several enterprise Linux distributions and appears in some NAS devices.
XFS journals metadata but keeps little information about deleted files, so undeleting by name is difficult; carving and analysis of leftover inode data are the usual approaches.
Its repair tool can discard metadata it can't reconcile, and it may ask to clear the log first, which throws away the most recent changes. With important data, image the volume and work on a copy.
Advanced · xfs · linux · nas · allocation groups
exFAT is built for flash media. It's the required format for SDXC cards and is common on USB drives and external disks shared between Windows and Mac, because both can read and write it without extra software. Unlike FAT32, it handles very large files.
exFAT has no journal, so pulling a card or drive out mid-write easily leaves the file system inconsistent. Many files are stored as one continuous run with a flag that lets exFAT skip the file allocation table entirely.
When a file is deleted, its directory entries are marked unused but mostly kept, which helps recovery. Fragmented files are harder, because their allocation chain may be lost. Always eject before unplugging.
Intermediate · exfat · sd card · flash drive · cross-platform
FAT32 is decades old and still everywhere: small USB drives, older cameras, car stereos, and devices that need maximum compatibility. It tracks files with a file allocation table, a chain of entries showing which cluster follows which, and keeps two copies of that table.
Its limits are well known: no single file larger than 4 GB, no journaling, and no permissions. Long videos are often split across several files because of this.
When a FAT32 file is deleted, the first character of its name is overwritten and its cluster chain is cleared. Files stored in one continuous run usually recover well; fragmented ones are harder, because the order of the pieces has to be worked out from their content.
Beginner · fat32 · file allocation table · usb drive · compatibility
ReFS is Microsoft's Resilient File System, used mainly on Windows Server, with Storage Spaces, and for backup repositories and virtualization storage. It uses copy-on-write for metadata, checksums to detect corruption, and, on mirrored or parity storage, can repair corrupted data automatically from a good copy.
ReFS doesn't use CHKDSK in the usual way and has fewer third-party recovery tools than NTFS. When a ReFS volume becomes unmountable, recovery means analyzing its tree structures directly.
Because ReFS is popular for backup storage, a damaged ReFS volume can mean losing production data and its backups at the same time. Keep at least one backup copy on different storage.
Advanced · refs · windows server · storage spaces · integrity
ZFS combines a volume manager and a file system. Drives are grouped into vdevs, such as mirrors or RAID-Z with single, double, or triple parity, and vdevs form a pool. Every block is checksummed and written copy-on-write, so ZFS can detect silent corruption and repair it from redundancy.
ZFS commits changes in transaction groups and keeps several recent root pointers called uberblocks. When a pool won't import, an older transaction group can sometimes be used to roll back to a consistent state, at the cost of the newest writes.
The danger is in panicked commands: forcing imports, replacing the wrong disk, or adding a drive to the wrong vdev can be hard to undo. If a pool fails, stop, record its layout and the error messages, and image the drives before trying anything with force options.
Advanced · zfs · raid-z · storage pool · checksums
Btrfs is a copy-on-write file system with snapshots, checksums, built-in RAID modes, and subvolumes. Many consumer NAS devices use it on top of their own RAID layer to offer snapshots and data-integrity checks.
Because metadata is never overwritten in place, older copies of tree roots often remain, and recovery tools can use them to read files from a damaged volume without modifying it.
Btrfs has a repair command, but its own documentation warns that it can cause further damage and should be a last resort. On a NAS, the problem may also be in the RAID layer underneath. Image the drives, then work through each layer in order.
Advanced · btrfs · nas · snapshots · copy-on-write
A partition table tells the computer where each partition starts and ends. The older MBR scheme lives in the first sector, allows four primary partitions, and can't address beyond about 2 TB on drives with standard 512-byte sectors. GPT, used by modern UEFI systems, supports far larger disks and many partitions, and keeps a checksummed backup copy of its header and table at the end of the disk.
If a partition table is damaged, the partitions and their data are usually still intact; the computer just doesn't know where they are. Rebuilding the table is often possible by locating each file system's boot sector or superblock.
Converting between MBR and GPT, or letting an operating system create a new table, can overwrite the old structures. Rebuild from evidence, not by guesswork.
Intermediate · mbr · gpt · partition table · uefi
A partition can vanish when someone deletes it in Disk Management or Disk Utility, when an installer repartitions a disk, or when the partition table is damaged. The drive then shows unallocated space, even though the file system inside may be untouched.
Deleting a partition usually just removes its entry from the table. The data stays put until a new partition is created and formatted over it, or new files are written.
Don't create a new partition or volume to get the space back. Recovery tools and labs can scan for the old file system's signatures and restore the partition entry exactly where it was.
Intermediate · partition recovery · deleted partition · disk management · unallocated space
Formatting a drive by mistake, such as choosing the wrong disk in an installer or in Disk Utility, is common and often recoverable. A quick format writes a fresh, empty file system but leaves most of the old file contents in place.
What decides the outcome is what happened next. If nothing was written afterward, the old files, and often their folder structure, can be recovered. If a new operating system was installed or files were copied in, those overwrote parts of the old data. On SSDs, formatting usually triggers TRIM, which can erase the old data quickly.
Stop using the drive immediately, don't reformat it back to the old file system, and recover to a different drive.
Beginner · accidental format · quick format · recovery · overwrite
When Disk Management shows a disk as Unknown and Not Initialized, Windows couldn't read a valid partition table. That happens when the first sectors are damaged, when a failing drive can't read them, when an external drive's USB adapter presents the disk differently than before, or when the disk uses a layout Windows doesn't understand.
Windows then offers to initialize it. Initializing writes a brand-new, empty partition table and can make the old partitions much harder to find.
Close the prompt. If the drive came out of an external enclosure, put it back. If it's making unusual noises or responds slowly, treat it as failing hardware. Otherwise, a careful partition recovery can rebuild the original table.
Intermediate · disk not initialized · windows · partition table · disk management
Every file system starts with a small structure describing itself: block size, where the main tables are, and how big the volume is. On NTFS and FAT it's the boot sector; on ext4 and similar systems it's the superblock; APFS and ZFS use their own container and label structures.
If that first structure is damaged, the whole volume looks blank or unformatted even though everything else is intact. Most file systems keep backups: NTFS at the end of the volume, ext file systems in several block groups, GPT at the end of the disk, and ZFS in labels at the start and end of each device.
Recovery often begins by finding a good backup copy and using it to read the volume, without writing anything to the original until the copy has proven correct.
Advanced · superblock · boot sector · backup · file system metadata
A journaling file system writes down what it's about to change before making the change. If power fails mid-update, the journal lets it finish or roll back the change instead of leaving half-written structures. NTFS, ext4, HFS+, and XFS all journal their metadata.
The journal is also a recovery and forensic resource. It can hold recent copies of metadata, including records of files that were just deleted, which tools can use to restore names and locations.
Journaling protects structure, not contents. A file that was being written when power failed can still be incomplete or corrupt, even when the file system itself is consistent.
Intermediate · journaling · ntfs · ext4 · crash recovery
Linux's Logical Volume Manager lets several drives or partitions form a volume group, which is then divided into logical volumes that hold file systems. Volumes can grow, span drives, and take snapshots.
LVM keeps its layout in metadata on each physical volume, and Linux systems usually save text backups of that metadata too. If a disk is removed, wiped, or fails, the volumes that used space on it become incomplete.
Recovery means restoring the metadata to match the original layout and imaging every drive that was part of the group. A logical volume that spanned a failed drive can often be partially recovered even if that drive cannot.
Advanced · lvm · linux · logical volume · volume group
Some iMacs and Mac minis shipped with a Fusion Drive: a small SSD and a hard drive combined into one volume, with macOS moving frequently used data to the SSD. To the user it's one disk; underneath, files can be split across both devices.
If either device fails, the combined volume breaks. Recovery means reading both, assembling the volume as macOS did, and only then reaching the files.
Don't split a Fusion Drive into separate disks or erase one half to fix it. Hybrid drives, or SSHDs, which combine a flash cache and platters in a single drive, are simpler: the drive manages the cache internally, and recovery usually centers on the platters.
Advanced · fusion drive · mac · sshd · hybrid drive
Windows has two ways to combine drives at the operating system level. Older dynamic disks keep a small database describing spanned, striped, or mirrored volumes. Storage Spaces, the modern replacement, pools drives and carves virtual disks out of the pool with simple, mirror, or parity layouts, spreading data in slabs across the drives.
Both depend on metadata stored on every member disk. If a drive fails or the metadata is damaged, the pool or volume may go offline, and individual drives show up as unknown or unusable when connected elsewhere.
Never format or initialize one member of a pool to test it. Keep all drives together and labeled, and the layout can be reassembled from the metadata and data patterns.
Advanced · dynamic disks · storage spaces · windows · spanned volume
A typical Windows or Mac disk has several partitions you never see in File Explorer or Finder: an EFI system partition with boot files, a recovery partition with repair tools, and sometimes a manufacturer's restore partition.
These are useful for repairing a computer, but a manufacturer's factory restore wipes your data partition and puts back the original software. It's not a way to recover files.
When a drive is imaged for recovery, every partition is captured. Your documents almost always live on the largest partition, and that's where recovery focuses.
Beginner · efi partition · recovery partition · hidden partition · windows