Linear Tape File System (LTFS)
Scope
LTFS is the open, self-describing file system that makes a tape mountable. The exam tests its on-tape structure, lifecycle commands, performance model, and where it is emphatically not a general-purpose file system.
Core rules
- LTFS uses the two-partition layout introduced with LTO-5: a small index partition and a large data partition.
- The index is an XML document holding file names, sizes, timestamps, extended attributes and extents. An extent maps a logical file offset to a tape block range, so one file can span non-contiguous regions.
- Mount reads the newest index; unmount or an explicit sync writes a new index generation and leaves earlier ones as fallback. Key LTFS attributes (format version, index generation) are mirrored in cartridge MAM so a drive identifies LTFS media quickly.
- The reference tools are mkltfs to format, ltfs to mount, an unmount to close, and ltfsck to check. ltfsck can rebuild a damaged index by scanning the data partition - slow, but possible precisely because the format is open and documented.
- Writes are append-only. Overwriting a file writes new extents elsewhere; deleting frees nothing until reformat. Small random writes are the worst case, because each sync costs an index write and a reposition.
Worked example
Copying a 100 GB file: LTFS streams extents into the data partition, then at unmount writes a new index of a few hundred kilobytes into the index partition after the previous index. If that index is corrupt, ltfsck scans the data partition and reconstructs the file list from the extent records - the data was never at risk, only the map to it.
Traps
- Treating LTFS as POSIX with journaling, locking and instant random writes.
- Forgetting that a WORM cartridge forces append-only behaviour.
- Confusing the LTFS format compression setting with the drive's actual ALDC compression.
- Expecting deleted files to return capacity.