Hi guys, I'm developing a submission platform for Flink jobs deployed in containers using Flink client for submission. I found that the Flink client by default uses the default JVM heap setting (with an initial heap size of 1/64 host memory, and a max heap size of 1/4 host memory). When the container memory is large, the client is configured to consume too much memory, and can easily causing container OOM when massive concurrent submissions take place.
I have two questions: 1. Is there a configuration to control the client heap size? 2. What's the suggested heap size to be used (I'm mostly using Per-Job mode where job graph is generated on the client side)? I'd appreciate a lot if anyone could give some insights on this. Thanks, Zhanghao Chen
