for question 1:
1, the basic principle is to load the context, of spring before execution in order to avoid repeated reconstruction, you can use the simple interest mode
2. For source, sink and transformation, you can use the relevant function, that inherits AbstractRichFunction to initialize spring context
3. For the destruction of spring context, register a shutdown hook to destroy

for problem 2:
see how to use timed tasks. If you only need to be similar to schedule, in ordinary spring, then problem 1 can be solved. If you want to do timing in source or sink, you probably have to implement it according to your own requirements, such as using cron4j

http://www.54tianzhisheng.cn/.

there are also many good Flink blogs here, which are worth recommending!

the corresponding code repository is located in https://github.com/zhisheng17.