Identifying Invalid CSV Files in Error Table Data
If a CSV file contains invalid formatting, the rawdata field in the error table can contain several combined rows. For example, if a closing quote for a specific field is missing, all the following newlines are treated as embedded newlines. When this happens, HAWQ stops parsing a row when it reaches 64K, puts that 64K of data into the error table as a single row, resets the quote flag, and continues. If this happens three times during load processing, the load file is considered invalid and the entire load fails with the message “
or more rows”. See Escaping in CSV Formatted Files for more information on the correct use of quotes in CSV files.