What causes "ORC file is empty: " in prestodb/presto?

Remove the empty file: hadoop fs -rm

prestodb/presto · error · PrestoException

HIVE_BAD_DATA

HIVE_BAD_DATA

Error message

ORC file is empty:

What it means

DwrfSelectivePageSourceFactory.createPageSource performs the same zero-length-file check for selective (row-level filtered) ORC scans, throwing PrestoException(HIVE_BAD_DATA) when fileSplit.getFileSize() == 0. After this check it also validates the rowID partition component before building the reader.

Source

Thrown at presto-hive/src/main/java/com/facebook/presto/hive/orc/DwrfSelectivePageSourceFactory.java:118

            List<HiveColumnHandle> columns,
            Map<Integer, String> prefilledValues,
            Map<Integer, HiveCoercer> coercers,
            Optional<BucketAdaptation> bucketAdaptation,
            List<Integer> outputColumns,
            TupleDomain<Subfield> domainPredicate,
            RowExpression remainingPredicate,
            DateTimeZone hiveStorageTimeZone,
            HiveFileContext hiveFileContext,
            Optional<EncryptionInformation> encryptionInformation,
            boolean appendRowNumberEnabled,
            Optional<byte[]> rowIDPartitionComponent)
    {
        if (!OrcSerde.class.getName().equals(storage.getStorageFormat().getSerDe())) {
            return Optional.empty();
        }

        if (fileSplit.getFileSize() == 0) {
            throw new PrestoException(HIVE_BAD_DATA, "ORC file is empty: " + fileSplit.getPath());
        }

        checkRowIDPartitionComponent(columns, rowIDPartitionComponent);

        return Optional.of(createOrcPageSource(
                session,
                DWRF,
                hdfsEnvironment,
                configuration,
                fileSplit,
                columns,
                prefilledValues,
                coercers,
                bucketAdaptation,
                outputColumns,
                domainPredicate,
                remainingPredicate,
                false,

View on GitHub (pinned to 55bb57d202)

Solutions

Remove the empty file: hadoop fs -rm <empty-file-path>
Re-run the producer job to write real data, then refresh the partition
Verify table location contents: hadoop fs -ls -R <table-location> | grep for 0-size files

Example fix

// before
SELECT * FROM t WHERE ds = '2026-01-01'; -- ORC file is empty: /warehouse/t/ds=2026-01-01/f

// after
-- hadoop fs -rm /warehouse/t/ds=2026-01-01/f
-- ALTER TABLE t DROP PARTITION (ds='2026-01-01'); then re-add after data lands
SELECT * FROM t WHERE ds = '2026-01-01';

Defensive patterns

Strategy: validation

Validate before calling

# verify selective-scan table location has no empty files
hadoop fs -ls -R /warehouse/t/ | awk '$5 == 0 {print $NF}'

Try / catch

try {
    return selectiveScan(sql, filters);
} catch (PrestoException e) {
    if (e.getErrorCode().getName().equals("HIVE_BAD_DATA")
        && e.getMessage().startsWith("ORC file is empty:")) {
        purgeEmptyFiles(tableLocation);
        return selectiveScan(sql, filters);
    }
    throw e;
}

Prevention

Guarantee streaming writers flush initial data or close/delete the file on failure
Register partitions only after writer commit succeeds
Alert on zero-byte files in actively queried partitions

When it happens

Trigger: Selective scan (with pushed-down filters or rowId usage) on an ORC/DWRF table where the assigned split's file is 0 bytes.

Common situations: Empty files from failed/partial writes; tables ingested by streaming writers that create the file but crash before the first flush; stale partitions pointing at cleaned directories.

Related errors

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/558f4e18afc231ac. Report an issue: GitHub.