apache/cassandra · error · MarshalException

The data cannot be deserialized as a set

Error message

The data cannot be deserialized as a set

What it means

SetSerializer.deserialize reads a 4-byte signed element count; a negative count cannot represent a valid set, so deserialization is aborted. This typically indicates bit-level corruption of the count field or reading bytes with the wrong layout/type. The guard prevents allocating a nonsensical collection.

Source

Thrown at src/java/org/apache/cassandra/serializers/SetSerializer.java:103

            if (!accessor.isEmptyFromOffset(input, offset))
                throw new MarshalException("Unexpected extraneous bytes after set value");
        }
        catch (BufferUnderflowException | IndexOutOfBoundsException e)
        {
            throw new MarshalException("Not enough bytes to read a set");
        }
    }

    @Override
    public <V> Set<T> deserialize(V input, ValueAccessor<V> accessor)
    {
        try
        {
            int n = readCollectionSize(input, accessor);
            int offset = sizeOfCollectionSize();

            if (n < 0)
                throw new MarshalException("The data cannot be deserialized as a set");

            // If the received bytes are not corresponding to a set, n might be a huge number.
            // In such a case we do not want to initialize the set with that initialCapacity as it can result
            // in an OOM when add is called (see CASSANDRA-12618). On the other hand we do not want to have to resize
            // the set if we can avoid it, so we put a reasonable limit on the initialCapacity.
            Set<T> l = new LinkedHashSet<>(Math.min(n, 256));

            for (int i = 0; i < n; i++)
            {
                V value = readNonNullValue(input, accessor, offset);
                offset += sizeOfValue(value, accessor);
                elements.validate(value, accessor);
                l.add(elements.deserialize(value, accessor));
            }
            if (!accessor.isEmptyFromOffset(input, offset))
                throw new MarshalException("Unexpected extraneous bytes after set value" + l + "," + accessor.toHex(input));
            return l;
        }

View on GitHub (pinned to 88fd0f6a0e)

Solutions

1. Verify the column type matches the stored value (DESCRIBE table; check type metadata).
2. Re-read the value; if reproducible, treat the cell as corrupt and rewrite it (scrub/repair).
3. Ensure the buffer position starts at the value's beginning, not mid-value (a misaligned position flips the sign bit).
4. Catch MarshalException in read paths and skip/log the corrupt value.

Example fix

// before
Set<Integer> tags = setSerializer.deserialize(cellValue, ByteBufferAccessor.instance);
// after
try {
    Set<Integer> tags = setSerializer.deserialize(cellValue.duplicate(), ByteBufferAccessor.instance);
} catch (MarshalException e) {
    logger.warn("Corrupt set cell ignored: {}", e.getMessage());
}

// Java
if (buf.remaining() >= 4) {
    int n = ByteBufferUtil.toInt(buf.duplicate());
    if (n < 0) throw new IllegalStateException("corrupt set count");
}

try {
    Set<T> s = setSerializer.deserialize(buf.duplicate(), ByteBufferAccessor.instance);
} catch (MarshalException e) {
    log.warn("Corrupt set cell: {}", e.getMessage());
}

When it happens

Trigger: Deserializing a buffer whose first 4 bytes, interpreted as an int, are negative — e.g. corrupt storage bytes, wrong column type, or reading a non-set value as a set.

Common situations: SSTable corruption; reading data written by a tool that does not use Cassandra's collection encoding; type confusion between column metadata and stored bytes.

Understand the failure class

Background: "Invalid ... format", "must be in format X", "does not look like a ..." — invalid argument format errors across CLI tools and libraries — this error's family across 17 libraries.

Related errors

• The data cannot be deserialized as a list
• Not enough bytes to read a set
• Unexpected extraneous bytes after set value
• Not enough bytes to read size of %dth field %s
• Not enough bytes to read %dth field %s

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/b83474886cf05b13. Report an issue: GitHub.