Github user suez1224 commented on the issue: https://github.com/apache/flink/pull/5867 Hi @twalthr , IMO, it's better to seperate the configuration of SQL client from the configuration of the Flink job. From the user's perspective, many of them probably do not need to know or use SQL client, putting a SQL client section in flink-conf.yaml will just confuse the users because they might think that's part of their flink job configuration. I suggest, as you already mentioned, we can either 1) add some variable in the sql-client.sh to allow the advanced users to modify the JVM ARGS within the script or 2) add a sql-client-conf.yaml to allow advanced user to edit it. Assuming that it is uncommon to modify the JVM options of the sql client, I think both options are fine, but I slightly prefer option 2 since sql-client.sh is not supposed to modified by the users.
---