Cassandra SSTable recovery

A corrupt Cassandra SSTable should be preserved before scrub discards unreadable rows or partitions.

Cassandra SSTables are immutable files with data, index, summary, filter, statistics and compression components. A checksum failure, truncated component or storage fault can affect one SSTable while other replicas still hold overlapping data.

Understand the symptom

Recovery compares immutable SSTable contents with replication, repair and tombstone history.

WHY

What is happening

Scrub can rewrite readable partitions and skip corrupt content, which may be useful only after protected copies and replica comparison. Recovery must account for consistency level, replication factor, token ownership, tombstones, repair timestamps and commit-log coverage before treating another node as complete.

What you may see

Symptoms that change the recovery path.

01

CorruptSSTableException

A read, compaction or startup task detects an unreadable component.

02

Checksum mismatch

Compressed chunks or file contents do not match stored checksums.

03

Compaction repeatedly fails

The same SSTable blocks maintenance and can keep disk usage growing.

04

One replica returns different data

Corruption, missed repair or tombstone history creates disagreement.

Protect the source

Actions that can make the situation worse.

01

Scrub may skip damaged data

The rewritten SSTable can open while silently omitting unreadable partitions.

02

Repair can propagate an incomplete view

A poorly chosen source can stream missing or deleted state across replicas.

03

Deleting an SSTable removes unique history

Replication does not guarantee every row and tombstone exists elsewhere at the same version.

Do this now

Safe steps before evaluation.

01

Stop compaction and repair

Pause automation that rewrites, deletes or streams the affected SSTables.

02

Identify every component

Record keyspace, table ID, generation, file list, checksums, node and token ranges.

03

Preserve cluster context

Copy data files, commit logs, snapshots, schema, system tables and nodetool status.

04

Check storage health

Capture filesystem, kernel, RAID and device errors before more reads.

NEPFOSYS process

How a data-first evaluation proceeds.

01

Replica and token mapping

NEPFOSYS compares affected ranges, SSTable generations, snapshots and other replica copies.

02

Copy-based extraction

Readable partitions and components are evaluated without rewriting originals.

03

Logical reconciliation

Recovered rows, timestamps and tombstones are validated before any cluster repair plan.

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.

Will nodetool scrub repair a corrupt SSTable?

It can rewrite readable content and skip corruption, but skipped data may be lost. Use it only after preserving the original and comparing replicas.

Can I delete the bad SSTable and run repair?

That may be viable only when another verified replica contains the required data and tombstone history. Deletion first removes the chance to inspect unique content.

Are Cassandra snapshots enough?

Snapshots preserve SSTable links at a point in time, but schema, commit logs, replica context and off-node copies are also important.

DFW local · Nationwide mail-in

Protect the device. Start with the facts.

Start my recovery case