我们最开始发现这个现象的时候也有些惊讶，不过后来想了一下感觉也是合理的。
因为双流Join的时间范围有可能会比较大，比如 A流 在 B流的[-10min, +10min]，那这样的话， A流来一条数据，可能会join到几分钟之前的数据，而此时的watermark其实已经大于了那条数据的事件时间。 我个人感觉，这应该就是在更实时的产生Join结果和导致数据时间晚于watermark之间，需要有一个balance。 现在默认实现是选择了更加实时的产生结果。当然还有另外一种实现思路，就是保证watermark不会超过数据时间， 那样的话，Join结果的产生就会delay，或者需要修改watermark逻辑，让watermark一定要小于当前能join到的数据 的时间最早的那个。 元始(Bob Hu) <[email protected]> 于2020年7月5日周日 下午8:48写道： > 谢谢您的解答。感觉flink这个机制有点奇怪呢 > > > ------------------ 原始邮件 ------------------ > *发件人:* "Benchao Li"<[email protected]>; > *发送时间:* 2020年7月5日(星期天) 中午11:58 > *收件人:* "元始(Bob Hu)"<[email protected]>; > *抄送:* "user-zh"<[email protected]>; > *主题:* Re: flink interval join后按窗口聚组问题 > > 回到你的问题，我觉得你的观察是正确的。Time interval join产生的结果的确是会有这种情况。 > 所以如果用事件时间的time interval join，后面再接一个事件时间的window（或者其他的使用事件时间的算子，比如CEP等） > 就会有些问题，很多数据被作为late数据直接丢掉了。 > > 元始(Bob Hu) <[email protected]> 于2020年7月3日周五 下午3:29写道： > >> 您好，我想请教一个问题： >> flink双流表 interval join后再做window group是不是有问题呢，有些left join关联不上的数据会被丢掉。 >> 比如关联条件是select * from a,b where a.id=b.id and b.rowtime between a.rowtime >> and a.rowtime + INTERVAL '1' HOUR >> ，看源码leftRelativeSize=1小时，rightRelativeSize=0，左流cleanUpTime = rowTime + >> leftRelativeSize + (leftRelativeSize + rightRelativeSize) / 2 + >> allowedLateness + >> 1，左表关联不上的数据会在1.5小时后输出（右表为null），而watermark的调整值是Math.max(leftRelativeSize, >> rightRelativeSize) + >> allowedLateness，也就是1小时，那这样等数据输出的时候watermark不是比左表rowtime还大0.5小时了吗,后面再有对连接流做group >> by的时候这种右表数据为空的数据就丢掉了啊。 >> flink版本 1.10.0。 >> >> 下面是我的一段测试代码： >> >> import org.apache.commons.net.ntp.TimeStamp; >> import org.apache.flink.api.common.typeinfo.TypeInformation; >> import org.apache.flink.api.common.typeinfo.Types; >> import org.apache.flink.api.java.typeutils.RowTypeInfo; >> import org.apache.flink.streaming.api.TimeCharacteristic; >> import org.apache.flink.streaming.api.datastream.DataStream; >> import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment; >> import >> org.apache.flink.streaming.api.functions.AssignerWithPeriodicWatermarks; >> import org.apache.flink.streaming.api.functions.ProcessFunction; >> import org.apache.flink.streaming.api.functions.source.SourceFunction; >> import org.apache.flink.streaming.api.watermark.Watermark; >> import org.apache.flink.table.api.EnvironmentSettings; >> import org.apache.flink.table.api.Table; >> import org.apache.flink.table.api.java.StreamTableEnvironment; >> import org.apache.flink.table.functions.ScalarFunction; >> import org.apache.flink.types.Row; >> import org.apache.flink.util.Collector; >> import org.apache.flink.util.IOUtils; >> >> import java.io.BufferedReader; >> import java.io.InputStreamReader; >> import java.io.Serializable; >> import java.net.InetSocketAddress; >> import java.net.Socket; >> import java.sql.Timestamp; >> import java.text.SimpleDateFormat; >> import java.util.ArrayList; >> import java.util.Date; >> import java.util.List; >> >> public class TimeBoundedJoin { >> >> public static AssignerWithPeriodicWatermarks<Row> getWatermark(Integer >> maxIdleTime, long finalMaxOutOfOrderness) { >> AssignerWithPeriodicWatermarks<Row> timestampExtractor = new >> AssignerWithPeriodicWatermarks<Row>() { >> private long currentMaxTimestamp = 0; >> private long lastMaxTimestamp = 0; >> private long lastUpdateTime = 0; >> boolean firstWatermark = true; >> // Integer maxIdleTime = 30; >> >> @Override >> public Watermark getCurrentWatermark() { >> if(firstWatermark) { >> lastUpdateTime = System.currentTimeMillis(); >> firstWatermark = false; >> } >> if(currentMaxTimestamp != lastMaxTimestamp) { >> lastMaxTimestamp = currentMaxTimestamp; >> lastUpdateTime = System.currentTimeMillis(); >> } >> if(maxIdleTime != null && System.currentTimeMillis() - >> lastUpdateTime > maxIdleTime * 1000) { >> return new Watermark(new Date().getTime() - >> finalMaxOutOfOrderness * 1000); >> } >> return new Watermark(currentMaxTimestamp - >> finalMaxOutOfOrderness * 1000); >> >> } >> >> @Override >> public long extractTimestamp(Row row, long >> previousElementTimestamp) { >> Object value = row.getField(1); >> long timestamp; >> try { >> timestamp = (long)value; >> } catch (Exception e) { >> timestamp = ((Timestamp)value).getTime(); >> } >> if(timestamp > currentMaxTimestamp) { >> currentMaxTimestamp = timestamp; >> } >> return timestamp; >> } >> }; >> return timestampExtractor; >> } >> >> public static void main(String[] args) throws Exception { >> StreamExecutionEnvironment bsEnv = >> StreamExecutionEnvironment.getExecutionEnvironment(); >> EnvironmentSettings bsSettings = >> EnvironmentSettings.newInstance().useBlinkPlanner().inStreamingMode().build(); >> StreamTableEnvironment bsTableEnv = >> StreamTableEnvironment.create(bsEnv, bsSettings); >> bsEnv.setParallelism(1); >> bsEnv.setStreamTimeCharacteristic(TimeCharacteristic.EventTime); >> >> >> // DataStream<Row> ds1 = bsEnv.addSource(sourceFunction(9000)); >> SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss"); >> List<Row> list = new ArrayList<>(); >> list.add(Row.of("001",new Timestamp(sdf.parse("2020-05-13 >> 00:00:00").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 00:20:00").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 00:40:00").getTime()), 100)); >> list.add(Row.of("002",new Timestamp(sdf.parse("2020-05-13 >> 01:00:01").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 02:20:00").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 02:30:00").getTime()), 100)); >> list.add(Row.of("003",new Timestamp(sdf.parse("2020-05-13 >> 02:00:02").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 02:20:00").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 02:40:00").getTime()), 100)); >> list.add(Row.of("004",new Timestamp(sdf.parse("2020-05-13 >> 03:00:03").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 03:20:00").getTime()), 100)); >> list.add(Row.of("000",new Timestamp(sdf.parse("2020-05-13 >> 03:40:00").getTime()), 100)); >> list.add(Row.of("005",new Timestamp(sdf.parse("2020-05-13 >> 04:00:04").getTime()), 100)); >> DataStream<Row> ds1 = bsEnv.addSource(new SourceFunction<Row>() { >> @Override >> public void run(SourceContext<Row> ctx) throws Exception { >> for(Row row : list) { >> ctx.collect(row); >> Thread.sleep(1000); >> } >> >> } >> >> @Override >> public void cancel() { >> >> } >> }); >> ds1 = ds1.assignTimestampsAndWatermarks(getWatermark(null, 0)); >> ds1.getTransformation().setOutputType((new RowTypeInfo(Types.STRING, >> Types.SQL_TIMESTAMP, Types.INT))); >> bsTableEnv.createTemporaryView("order_info", ds1, "order_id, >> order_time, fee, rowtime.rowtime"); >> >> List<Row> list2 = new ArrayList<>(); >> list2.add(Row.of("001",new Timestamp(sdf.parse("2020-05-13 >> 01:00:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 01:20:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 01:30:00").getTime()))); >> list2.add(Row.of("002",new Timestamp(sdf.parse("2020-05-13 >> 02:00:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 02:20:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 02:40:00").getTime()))); >> // list2.add(Row.of("003",new Timestamp(sdf.parse("2020-05-13 >> 03:00:03").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 03:20:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 03:40:00").getTime()))); >> list2.add(Row.of("004",new Timestamp(sdf.parse("2020-05-13 >> 04:00:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 04:20:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 04:40:00").getTime()))); >> list2.add(Row.of("005",new Timestamp(sdf.parse("2020-05-13 >> 05:00:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 05:20:00").getTime()))); >> list2.add(Row.of("111",new Timestamp(sdf.parse("2020-05-13 >> 05:40:00").getTime()))); >> DataStream<Row> ds2 = bsEnv.addSource(new SourceFunction<Row>() { >> @Override >> public void run(SourceContext<Row> ctx) throws Exception { >> for(Row row : list2) { >> ctx.collect(row); >> Thread.sleep(1000); >> } >> >> } >> >> @Override >> public void cancel() { >> >> } >> }); >> ds2 = ds2.assignTimestampsAndWatermarks(getWatermark(null, 0)); >> ds2.getTransformation().setOutputType((new RowTypeInfo(Types.STRING, >> Types.SQL_TIMESTAMP))); >> bsTableEnv.createTemporaryView("pay", ds2, "order_id, pay_time, >> rowtime.rowtime"); >> >> Table joinTable = bsTableEnv.sqlQuery("SELECT a.*,b.order_id from >> order_info a left join pay b on a.order_id=b.order_id and b.rowtime between >> a.rowtime and a.rowtime + INTERVAL '1' HOUR where a.order_id <>'000' "); >> >> bsTableEnv.toAppendStream(joinTable, Row.class).process(new >> ProcessFunction<Row, Object>() { >> @Override >> public void processElement(Row value, Context ctx, >> Collector<Object> out) throws Exception { >> SimpleDateFormat sdf = new SimpleDateFormat("yyyy-MM-dd >> HH:mm:ss.SSS"); >> System.err.println("row:" + value + ",rowtime:" + >> value.getField(3) + ",watermark:" + >> sdf.format(ctx.timerService().currentWatermark())); >> } >> }); >> >> bsTableEnv.execute("job"); >> } >> } >> >> > > -- > > Best, > Benchao Li > -- Best, Benchao Li
