What causes "Internal file %s doesn't exist" in prestodb/presto?

List available columns in the segment metadata and query only existing ones

prestodb/presto · error · PrestoException

DRUID_SEGMENT_LOAD_ERROR

DRUID_SEGMENT_LOAD_ERROR

Error message

Internal file %s doesn't exist

What it means

Thrown when a requested internal (smoosh) column name is absent from the parsed metadata of a Druid smooshed segment. The column-to-chunk mapping (columnSmoosh) has no entry for the name, so the connector cannot locate the data inside the segment files.

Source

Thrown at presto-druid/src/main/java/com/facebook/presto/druid/segment/SmooshedColumnSource.java:62

    }

    @Override
    public int getVersion()
    {
        try {
            return ByteBuffer.wrap(indexFileSource.readFile(VERSION_FILE_NAME)).getInt();
        }
        catch (IOException e) {
            throw new PrestoException(DRUID_SEGMENT_LOAD_ERROR, e);
        }
    }

    @Override
    public byte[] getColumnData(String name)
    {
        SmooshFileMetadata metadata = columnSmoosh.get(name);
        if (metadata == null) {
            throw new PrestoException(DRUID_SEGMENT_LOAD_ERROR, format("Internal file %s doesn't exist", name));
        }
        String fileName = makeChunkFileName(metadata.getFileCount());
        int fileStart = metadata.getStartOffset();
        int fileSize = metadata.getEndOffset() - fileStart;

        byte[] buffer = new byte[fileSize];
        try {
            indexFileSource.readFile(fileName, fileStart, buffer);
        }
        catch (IOException e) {
            throw new PrestoException(DRUID_SEGMENT_LOAD_ERROR, e);
        }
        return buffer;
    }

    private void loadSmooshFileMetadata()
    {
        try {

View on GitHub (pinned to 55bb57d202)

Solutions

List available columns in the segment metadata and query only existing ones
Re-fetch/refresh segment metadata so the descriptor matches the smoosh files
Check for Druid segment version skew; re-publish the segment
Fix the query or column mapping to use the correct column name

Example fix

// before
SmooshFileMetadata metadata = columnSmoosh.get(name);
if (metadata == null) { throw ... }
// after (caller-side guard)
if (!availableColumns.contains(columnName)) { return Optional.empty(); }

Defensive patterns

Strategy: validation

Validate before calling

// caller-side pre-check against segment metadata
Set<String> columns = columnNamesFromSegmentMetadata(segmentMetadata);
if (!columns.contains(columnName)) {
    throw new PrestoException(NOT_FOUND, "Column not in segment: " + columnName);
}

Type guard

function hasColumn(metadata, name) { return metadata != null && metadata.getColumns().hasOwnProperty(name); }

Try / catch

try {
    byte[] data = columnSource.getColumnData(name);
} catch (PrestoException e) {
    if (e.getErrorCode().getName().equals("DRUID_SEGMENT_LOAD_ERROR") && e.getMessage().contains("doesn't exist")) {
        return Optional.empty();
    }
    throw e;
}

Prevention

Always validate column names against segment metadata.json before reading
Keep the segment descriptor and smoosh files from the same segment version
Invalidate caches when segments are replaced or re-published
Log the available column set when this error occurs to speed diagnosis

When it happens

Trigger: getColumnData(name) is called with a column name that does not exist in the segment's metadata.json-derived smoosh index — e.g. querying a column not present in the Druid segment or a stale schema.

Common situations: Querying columns added after the segment was published, pointing at a mismatched segment version, typos in column names, schema drift between Druid datasource and cached segment descriptors.

Related errors

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/d149ecb258c19fbf. Report an issue: GitHub.