I\'m trying to upgrade from Spark 2.1 to 2.2. When I try to read or write a dataframe to a location (CSV or JSON) I am receiving this error:
Illegal pattern
Use commons-lang3-3.5.jar fixed the original error. I didn't check the source code to tell why but it is no surprising as the original exception happens at org.apache.commons.lang3.time.FastDatePrinter.parsePattern(FastDatePrinter.java:282). I also noticed the file /usr/lib/spark/jars/commons-lang3-3.5.jar (on an EMR cluster instance) which also suggest 3.5 is the consistent version to use.