HDF5 data recovery

An HDF5 file may contain readable groups or datasets even when metadata or the final write is incomplete.

HDF5 organizes groups, datasets, links, attributes, datatypes, chunk indexes, and raw data through metadata structures. A crash during writing can leave the container internally inconsistent.

Understand the symptom

HDF5 metadata maps named objects and chunked data to locations inside or outside the file.

WHY

What is happening

The HDF5 library flushes metadata and data during close; an interrupted writer can leave incomplete state. Superblocks, object headers, B-trees, heaps, chunk indexes, filters, external links, and raw data may be damaged separately.

What you may see

Symptoms that change the recovery path.

01

Unable to open file or bad superblock

The signature, superblock, or root metadata may be unreadable.

02

Object not found

A link or object header may be missing even when raw data remains.

03

Chunk read or filter error

A chunk, checksum, compression filter, or index may fail.

04

File is smaller than expected

The writer may have stopped before all metadata or data was committed.

Protect the source

Actions that can make the situation worse.

01

Repacking can omit unreadable objects

Preserve the source before any conversion.

02

Update mode adds metadata writes

Use read-only duplicates for inspection.

03

External links can be overlooked

Collect every referenced file and path.

Do this now

Safe steps before evaluation.

01

Copy the complete data set

Keep primary files, external raw data, linked HDF5 files, logs, and manifests.

02

Record expected schema

List groups, datasets, shapes, datatypes, chunking, filters, and attributes.

03

Preserve software context

Keep library version, plugins, writer code, environment, and crash timeline.

04

Image unstable storage

Protect the device before large recursive reads.

NEPFOSYS process

How a data-first evaluation proceeds.

01

Container and storage assessment

NEPFOSYS maps readable metadata, object references, chunks, filters, and source health.

02

Selective extraction

Recoverable datasets and attributes are copied to new containers without modifying the original.

03

Scientific validation

Names, shapes, datatypes, dimensions, sample values, attributes, and expected counts are compared.

A clear next step

Preserve the original device before another repair or recovery attempt.

Start a case with the exact symptoms, incident timeline, prior attempts, encryption details, and the files that matter most. NEPFOSYS will route the case to the appropriate DFW intake or nationwide mail-in path.

Start with a case review

Questions, answered

What customers ask before starting.

Can h5repack repair a corrupt file?

It can copy readable objects, but unreadable objects may be omitted. Run it only on a duplicate after inventory.

Does a bad superblock mean raw data is gone?

Not necessarily; accessibility depends on the location and extent of damage.

Why keep the writer code?

It documents schema, shapes, filters, external links, and expected data needed for validation.

DFW local · Nationwide mail-in

Protect the device. Start with the facts.

Start my recovery case