prestodb/presto · error · PrestoException

HIVE_UNKNOWN_ERROR

HIVE_UNKNOWN_ERROR

Error message

Error reading from %s at position %s. 

What it means

The final catch-all in HdfsOrcDataSource.readInternal: any exception that is neither BlockMissingException nor an IOException becomes PrestoException(HIVE_UNKNOWN_ERROR) prefixed with 'Error reading from %s at position %s. '. This preserves the original exception as the cause for later diagnosis.

Source

Thrown at presto-hive/src/main/java/com/facebook/presto/hive/orc/HdfsOrcDataSource.java:79

    {
        try {
            long readStart = System.nanoTime();
            inputStream.readFully(position, buffer, bufferOffset, bufferLength);
            stats.readDataBytesPerSecond(bufferLength, System.nanoTime() - readStart);
        }
        catch (PrestoException e) {
            // just in case there is a Presto wrapper or hook
            throw e;
        }
        catch (Exception e) {
            String message = format("Error reading from %s at position %s. ", this, position);
            if (e.getClass().getSimpleName().equals("BlockMissingException")) {
                throw new PrestoException(HIVE_MISSING_DATA, message, e);
            }
            if (e instanceof IOException) {
                throw new PrestoException(HIVE_FILESYSTEM_ERROR, message + e.getMessage(), e);
            }
            throw new PrestoException(HIVE_UNKNOWN_ERROR, message, e);
        }
    }
}

View on GitHub (pinned to 55bb57d202)

Solutions

1. Examine the full stack trace (cause) in the query failure to find the true root exception
2. Check worker logs and memory pressure; retry on a healthy cluster if transient
3. Verify HDFS delegation tokens / Kerberos credentials are valid for long-running queries
4. If reproducible, report with the cause stack trace to the Presto/Hive connector maintainers

Example fix

// before
SELECT * FROM t; -- HIVE_UNKNOWN_ERROR: Error reading from ... at position ...

// after (diagnose)
-- inspect cause stack trace in worker log; renew tokens / increase heap as indicated
SELECT * FROM t;

// precheck cluster health and worker memory before heavy scans
-- check HDFS delegation token validity for long queries (hdfs gettoken / job conf)

try {
    return query(sql);
} catch (PrestoException e) {
    if (e.getErrorCode().getName().equals("HIVE_UNKNOWN_ERROR")) {
        logRootCause(e.getCause()); // the original non-IO exception carries the diagnosis
        throw new RuntimeException("ORC read failed; see cause", e.getCause());
    }
    throw e;
}

When it happens

Trigger: A non-IOException (e.g., RuntimeException, OOM from a huge read buffer, UncheckedIOException variant, security exception) escapes the HDFS read call at the requested position.

Common situations: Presto/HDFS client bugs, out-of-memory on workers reading very large columns, Kerberos/token expiry surfacing as unchecked exceptions, misconfigured HDFS settings causing runtime failures.

Related errors

• HIVE_FILESYSTEM_ERROR
• HIVE_BAD_DATA
• HIVE_BAD_DATA
• HIVE_BAD_DATA
• HIVE_MISSING_DATA

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/6b5674f4af30d807. Report an issue: GitHub.