Clean-room engine · reads Oracle artefacts at the byte level · v4.19 · contains no Oracle software
reclaimDB · block-level Oracle extraction

Recover & extract Oracle data without the database.

Recover and extract supported Oracle tables directly from datafiles, Data Pump dumps and RMAN backup pieces. No running database or Oracle client required. Parallel export also supplies downstream systems from suitable files of running databases, standbys and offline snapshots.

✕ SQL connection not required ✕ Oracle license no SE/EE keys ✕ network path air-gapped OK
RECLAIMDB · BYTE-EXACT ENGINESRC 0x2000
DATAFILE DECODE TABLE RAW BLOCKS · UNORDEREDREASSEMBLED · BYTE-EXACT ✓ BLOCK 0x2000 · ROW 0
// From the lab

Latest from the lab.

The two latest posts from the reclaimDB blog.

// Capabilities

Oracle data recovery and extraction.

Choose your source. Each path has its own maturity and limits; the support matrix below keeps those visible.

Beta · datafile recovery reclaimDB · dbf-extract

Read Oracle datafiles. No instance required.

Extract supported tables directly from .dbf files. Metadata detection and parallel export are GA; datafile recovery as a whole is Beta.

  • BASIC / OLTP on 19c / 26ai; source-specific limits apply
  • Wallet TDE and point-in-time recovery: Beta, within validated forms
How it works → Running-database export benchmark →
GA · Data Pump reclaimDB · dmp-extract

Read Oracle dumps. Without Oracle.

Export supported Data Pump tables and LOBs without importing the dump into Oracle.

  • 11.2 / 12.2 / 19c / 26ai: documented source forms
  • HCC / KGC and classic exp: Experimental; encrypted dumps refused
How it works → 11g lab: CSV & BLOB images →
Beta · recovery / forensics use case · forensics & DR

Recover readable data from damaged files.

Extract surviving records from supported offline sources and record file hashes for integrity checks.

  • Corruption isolation: GA; datafile recovery: Beta
  • ASM scanning: on request; ASM file extraction: Experimental
How it works →
GA inventory · Beta recovery use case · backup access

Inspect supported RMAN pieces.

Read backup-piece inventory and extract supported data. Readability alone proves neither a complete backup set nor a successful restore.

  • Piece parsing and inventory: GA
  • Decryption, merge and direct input: Beta, with source-specific limits
How it works →
GA / Beta · source-dependent use case · audit access

Find records in offline sources.

Extract supported archive data for review without rebuilding a live Oracle environment.

  • Data Pump: GA; datafile recovery: Beta
  • Source and recovery limits still apply to targeted access
How it works →
Beta · datafile workflow use case · open archive

Turn a snapshot into open files.

Extract supported datafiles from an application-consistent snapshot or clone, away from production. Use the open files for archives and downstream systems.

  • Datafile recovery: Beta; supported output formats only
  • Mounted source files required; storage access is a separate step
How it works →
// The pipeline

Source → engine → open formats.

Supported source files are decoded into rows and exported to open formats. Decryption and reconstruction depend on the source; warnings and refused forms remain explicit.

RECLAIMDB · BYTE-EXACT ENGINESRC *.DBF · DECODE BLOCK 0x2000
Datafile *.dbf · block 0x2000 Data Pump *.dmp · magic 0x2815 RMAN piece 512-byte blocks Snapshot ASM · EBS · SnapMirror ENGINE CORE 01 decode 02 decrypt 03 reassemble 04 verify byte-exact · SHA-256 sealed SOURCES · NO ORACLE REQUIRED ENGINE VERIFIED OUTPUT
// 15 open formats CSVJSONNDJSONXMLSQLSQL*Loader PostgreSQLSQLiteXLSX ParquetAvroORCArrowDeltaIceberg
Sources · no Oracle required
Data Pump .dmp Datafiles .dbf RMAN backup pieces ASM raw / snapshots
reclaimDB engine
01decode
02decrypt
03reassemble
04verify
15 open formats · byte-exact
CSVJSONNDJSONXMLSQLSQL*Loader PostgreSQLSQLiteXLSX ParquetAvroORCArrowDeltaIceberg

Runs where your data stays.

A self-contained CLI for direct file extraction. Work with locally accessible files or mounted storage. No Oracle client or telemetry; automation stays on your host.

// Evidence

What it handles — and what it won't.

Data Pump extraction and RMAN inventory are GA. Datafile recovery and several advanced paths are Beta. Source version, stored form and the limits below determine support.

Full support matrix: GA, Beta, Experimental and refused forms
Supported

Verified & in scope

GA— regression-verified scope
+Metadata auto-detect from SYSTEM.dbf · block-level corruption isolation
+Multi-threaded export --parallel — deterministic, byte-exact merge
+Data Pump extraction — 13 writers · BasicFile / SecureFile inline and out-of-row LOBs
+LOB side-files with length and SHA-256 manifest · 100 MiB DMP BLOB coverage
+BASIC & OLTP compression on 19c / 26ai · DBF SecureFile inline LOBs, including COMPRESS LOW
+RMAN piece parsing & catalog inventory — set completeness requires an independent catalog or manifest
+TDE raw-key input · 15 DBF export formats, including PostgreSQL COPY files, Parquet, Avro, Delta and Iceberg
+DMP version scope: Oracle 11.2 / 12.2 / 19c / 26ai · JSON/OSON, VECTOR and BOOLEAN in their documented source formats
Beta— functional, bounded qualification
+Datafile recovery .dbf · row chaining / migration · IOT recovery within the documented shapes
+DBF BasicFile / SecureFile out-of-row LOBs · multi-extent maps · compressed BLOB streaming with --lob-dir
+TDE password wallet ewallet.p12 · portable auto-login cwallet.sso within the validated wallet scope
+RMAN Password-only / Dual password route (pre-23) · 19c Transparent / Dual wallet route · validated 26ai wallet-XTS forms; geometry limits apply
+RMAN L0 + L1 merge, ZSTD MEDIUM and direct piece input — externally asserted piece counts required
+RMAN archived-redo extraction — complete plain 19c sets, 512 / 4096-byte redo; broader forms refused
+Point-in-time recovery — single-thread scalar redo, INSERT / UPDATE / DELETE, gap detection; unsupported changes refuse
+Tiered consistency certificate — records separate guarantees and gaps; a committed subset is not a single-SCN snapshot
+OBJECT / VARRAY / NESTED TABLE / supported SDO forms — native 19c / 26ai DMP; bounded scalar-composite DBF support with proven metadata
+LONG / LONG RAW — native 19c / 26ai DMP; separately proven inline native26 DBF cells only
+Deleted-row extraction --include-deleted · logical Oracle DDL sidecars for DBF / Data Pump; omissions named explicitly
+Index-backed partial-row recovery & block repair into a new DBF copy · ASM disk scanning — Recovery & Forensics on request
+DMP multi-file inventory and plain row continuation · XMLTYPE binary forms within documented vocabulary / decoder limits
+Type inference without metadata — explicitly heuristic; --require-metadata refuses the fallback
Experimental— limited evidence / review pending
+HCC (Hybrid Columnar) · DMP KGC compression — explicit opt-in, provenance / qualification review pending
+ASM file extraction · classic exp / imp dumps — synthetic / limited fixture evidence
+11.2 BASIC / OLTP observed compression forms — exact witnessed cases, broader layouts refused
+Native26 datafile-TDE envelope and additional 19c TDE algorithms — provenance under review; validated 19c AES256 remains Beta
Refused / out‑of‑scope

Won't touch — by design

Refused— unsupported source forms
✕Encrypted Data Pump dumps — Password, Transparent and Dual refuse before row parsing; re-export ENCRYPTION=NONE
✕Column-level ENCRYPT USING · ANYDATA / REF · BFILE contents (locator only) · external / temporary tables
✕PITR with RAC multi-thread redo, unsupported change vectors or out-of-row LOB redo; OLTP-compressed full-row redo is not supported
✕RMAN KGC/BASIC compression, encrypted multi-piece groups and unvalidated encryption / piece geometries
✕LONG on older / unknown DBF versions or 11.2 / 12.2 DMP; chained DBF LONG remains unproved
✕Unresolved XMLTYPE dictionaries, unrecognised binary forms and decoder limits — no guessed or partial XML
✕Composite graphs, collection element types or storage mappings outside the proven native 19c / 26ai subset
✕Unknown DMP charsets refuse by default; Hebrew / Turkish / Arabic MSWIN decoders are available, with iconv round-trip evidence only
Limits— explicit boundaries
○A readable RMAN piece does not prove a complete backup set or a successful database restore
○18c / 21c / 23ai native dumps are best-effort with a warning; --strict-version-gate refuses them
○Logical DDL sidecars declare unsupported or omitted objects; they do not recreate an entire Oracle environment
○Overwritten deleted rows and non-indexed cells lost to corruption cannot be reconstructed by guessing
// Scope reviewed 4 November 2026 · unsupported data is named with structured refusals such as OL_REFUSE_*; warnings, partial results and missing evidence remain explicit. No blanket recovery or completeness guarantee.
// Practical answers

Oracle recovery and extraction questions.

Start with the source files you have and the result you need.

What is reclaimDB, and who is it for?

reclaimDB is a command-line tool for Oracle data recovery and extraction. It reads supported datafiles, Data Pump dumps and RMAN backup pieces directly, without requiring a running Oracle instance or client. Database administrators, recovery teams and data engineers use it to recover data, access archives and export tables for downstream systems.

Explore the use cases →
Which Oracle files and versions are supported?

Supported sources include Oracle datafiles (.dbf), Data Pump dumps (.dmp) and RMAN backup pieces. Data Pump extraction covers documented source forms from Oracle 11.2, 12.2, 19c and 26ai. Datafile and RMAN support has separate version, compression and encryption limits. The support matrix distinguishes verified, Beta, Experimental and refused forms.

Check the support matrix →
Can I read an Oracle dump without Oracle?

reclaimDB reads supported Data Pump dumps directly from disk and exports their tables into open formats. No import into a running Oracle database is required. Source versions, encryption and stored forms still determine what can be recovered.

See Data Pump extraction →
Can deleted Oracle rows be recovered without Flashback?

Deleted row images may remain in a datafile. The reclaimDB lab walkthrough shows extraction with --include-deleted, review and a small reimport. Overwritten row data cannot be reconstructed, and recovered values must be checked against the incident.

Read the recovery walkthrough →
Does reading an RMAN piece prove my backup is restorable?

No. Reading or extracting a supported piece is different from proving backup-set completeness or successfully restoring an Oracle database. Establish the required pieces independently and review the documented source-specific limits.

See RMAN backup access →
Does reclaimDB restore a complete Oracle database?

reclaimDB extracts supported tables and data into open formats. This is different from restoring a complete, operational Oracle database. Logical DDL sidecars describe supported objects and name omissions; they do not recreate an entire Oracle environment. Recovery depends on the available source files and their condition.

Review capabilities and limits →
Can I extract data while the Oracle database is running?

Yes, when the source files are in a supported state. In our published benchmark, Oracle 19c remained running while reclaimDB read its original datafiles; the eight source tablespaces were read-only. The benchmark demonstrates extraction from a running database in that setup. Consistency for tables changing during extraction is a separate requirement.

See the running-database benchmark →
Can I use a standby database or an offline snapshot?

Datafiles made available from a standby environment or an application-consistent offline snapshot can serve as sources for downstream extraction within the supported file scope. Storage access, source preparation and consistency must be established for that environment. An actively changing standby is not automatically a fixed, consistent view.

Explore standby and snapshot workflows →
Which downstream systems can use the extracted data?

Use the exported files in analytics, data platforms, data lakes and migration workflows. DBF extraction offers 15 output formats, including CSV, JSON, Parquet, Arrow and PostgreSQL COPY files. Format availability depends on the input path; loading the files into the target system is a separate step.

Explore the output formats →
What is planned for read consistency?

Read-consistent extraction is a planned extension for downstream data pipelines. The goal is to extract a defined, consistent view while source databases continue to operate. This is a roadmap item. Current guarantees depend on the source state and documented support limits; the existing consistency certificate does not establish general read consistency for changing tables.

Review current consistency limits →
How do I request an evaluation?

Email office@reclaimdb.com with the source files you have and what you need to recover or extract. Include the Oracle version if known, any damage, compression or encryption, and whether the files come from a running database, a standby, a snapshot or a backup. For downstream exports, also describe the target system and consistency requirements.

Request an evaluation →

Your data. Reclaimed.

Tell us which source files you have and what you need to recover or extract. Request an evaluation and a source-specific assessment.