Is this still an issue we need to address? Will close otherwise
Former user
April 15, 2016 at 4:39 PM
The failure is produced when a loading the metadata of a custom-idexed table, even if the Spark connector doesn't use the index, so maybe it should be considered a bug instead of an improvement.
Major
Regarding to CASSANDRA-10217 the PerRowSecondaryIndex creation query has changed in cassandra 3.X
From:
To:
but the Spark-cassandra-connector fails loading the table schema during a saveToCassandra operation looking for an empty fake-column name.
More info about the exception here