Hi - I’m thinking of using Flink SQL, and need help with my use case. Stateless operation (filtering) in high scale with 160+- kafka topics (output single topic). A single SQL that will hold the query for all the 160 topics. About 500~600 times a year it’s expected to edit the tables to fit latest DDL, and edit the query. This means quite a lot of restarts of the query. Do you think this fits more the DataStream API, or flink sql should be fine?

either api should be fine, so you can go with flink sql

Every restart will include an alter for all changed data-types. I much prefer flink sql, but I’m quite afraid about how much it will damage my latency for every restart. (but since it’s stateless maybe it will be faster than expected)