apache/cassandra · error · IndexOutOfBoundsException

The requested position exceeds the index length

Error message

The requested position exceeds the index length

What it means

BigTableKeyReader.indexPosition(long) seeks the Index.db reader to an arbitrary position. It validates the requested position against the index file length and throws IndexOutOfBoundsException('The requested position exceeds the index length') when the position lies past EOF, preventing a failed seek on a corrupt or stale index reference.

Source

Thrown at src/java/org/apache/cassandra/io/sstable/format/big/BigTableKeyReader.java:184

    public RowIndexEntry rowIndexEntry() {
        assert detailed;
        return rowIndexEntry;
    }

    public FileHandle indexFile() {
        return indexFile;
    }

    public long indexPosition()
    {
        return indexFileReader.getFilePointer();
    }

    public void indexPosition(long position) throws IOException
    {
        if (position > indexLength())
            throw new IndexOutOfBoundsException("The requested position exceeds the index length");
        indexFileReader.seek(position);
        key = null;
        keyPosition = 0;
        dataPosition = 0;
        advance();
    }

    public long indexLength()
    {
        return indexFileReader.length();
    }

    @Override
    public void reset() throws IOException
    {
        indexFileReader.seek(initialPosition);
        key = null;
        keyPosition = 0;

View on GitHub (pinned to 88fd0f6a0e)

Solutions

1. Run `nodetool scrub` on the table to rebuild/skip corrupt index structures.
2. Remove the suspect SSTable and run repair to restore good data.
3. Validate Index.db/Summary.db with `sstableverify`/`sstablemetadata` offline tools.
4. Check for disk-level truncation and hardware issues; verify file sizes against backup copies.

Example fix

// before: blindly seeking into index
keyReader.indexPosition(entry.position); // throws if beyond EOF
// after: guard the seek
if (entry.position <= keyReader.indexLength())
    keyReader.indexPosition(entry.position);
else
    handleCorruptIndex(entry);

// Java: bounds-check before seek
if (position >= 0 && position <= keyReader.indexLength())
    keyReader.indexPosition(position);
else
    logCorruptIndexEntry(position);

boolean validIndexPos = (position >= 0 && position <= indexLength());

try {
    keyReader.indexPosition(pos);
} catch (IndexOutOfBoundsException e) {
    logger.warn("Index position {} beyond index length {} — index corrupt", pos, keyReader.indexLength());
}

When it happens

Trigger: A caller (e.g. KeyReader reset/seek, scrubber or verifier following an index entry) supplies a position greater than indexLength() — typically from a corrupted Index.db/Summary.db entry, a stale cached first/last index bound, or a file truncated concurrently.

Common situations: Corrupt or truncated Index.db after crash/disk failure; summary entries pointing beyond the index after an interrupted compaction; reading an SSTable whose file was replaced/truncated underneath a live reader.

Related errors

• Invalid negative chunk index %d with position %d
• The requested position exceeds the index length
• Failed to read partition index
• Key from data file (%s) does not match key from index file (
• First partition does not match index

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/0390fe1b742576e4. Report an issue: GitHub.