Flink多并行度下watermark生成规则

一、多并加粗样式行度流的watermarks

注意：多并行度的情况下，watermark对齐会取所有channel最小的watermark

二、单个流多并行度
测试代码如下：

public class MutiParaWatermarkTest {
    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setMaxParallelism(2);
        env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime);
        env.getConfig().setAutoWatermarkInterval(200l);
        //1565488800000-2019/8/11 10:00:00
        DataStream<String> source = env.fromElements("1,a,1565488800000", "2,a,1565488801000", "3,b,1565488818000", "4,a,1565488810000", "5,b,1565488813000", "6,c,1565488812000", "7,c,1565488816000", "8,d,1565488822000", "9,c,1565488821000", "10,a,1565488831000");
        source.flatMap(new FlatMapFunction<String, Tuple3<String, String, Long>>() {
            @Override
            public void flatMap(String value, Collector<Tuple3<String, String, Long>> out) throws Exception {
                String[] strs = value.split(",");
                out.collect(Tuple3.of(strs[0], strs[1], Lo