Does persisting catalog/table information in Hive ...
# troubleshooting
v
Does persisting catalog/table information in Hive really make sense for non-interactive flink-sql workloads ? I do not think this is required. Any thoughts on this ?
d
what’ s a non-intereactive flink-sql workload. How do you define that?
v
A non-interactive flink-sql workload is where a user would submit a sql and it would run in the background. Think
INSERT INTO
queries.
d
How about using rocksdb?
v
I do not think I follow. Isn't that just a state backend.
d
Yes, but its a lot simpler than Hive so it’s not the same effort.
v
Cannot find any doc where rocksdb can be used as a catalog store.
d
there are some limits on what catalog methods it implements however.
Here is another resource on this topic: https://www.decodable.co/blog/catalogs-in-flink-sql-a-primer