I am running a SPARK JDBC process to extract data from Teradata. The target files are written into HDFS as ORC. I use the jdbcDF.write.format("orc").save(file) to save files onto HDFS. I run 8 parallel threads using different where clauses on the same table. Most of the times the process succeeds and some times 4-5 out of 8 parallel threads fail with the above error. But still the files are committed into HDFS and the counts match with source. However the temporary files renaming says failed but a permanent file is created and _SUCCESS file is not present in these failed target folders.
... View more