flink 1.12.2 所有事件都被删除了

Question

我的 flink 管道如下所示

FlinkKafkaConsumerBase kafkaConsumer = new FlinkKafkaConsumer<>(topic, new DeserializationSchema(),props);

kafkaSource = env.addSource(kafkaConsumer).filter(<>);
WatermarkStrategy<GenericMetricV2> watermarkStrategy = WatermarkStrategy
                .<GenericMetricV2>forBoundedOutOfOrderness(Duration.ofSeconds(900))
                .withTimestampAssigner((metric, timestamp) -> {
                    logger.info("ETS: mts: {}, ts: {}", metric.metricPoint.timeInstant, timestamp);
                    return metric.metricPoint.timeInstant;
                });

metricStream = kafkasource
                        .process(<>)
                        .assignTimestampsAndWatermarks(watermarkStrategy)
                        .transform("debugFilter", TypeInformation.of(<>), new StreamWatermarkDebugFilter<>("Op"))
                        .filter(<>)
                        .map(<>)
                        .flatMap(<>)
                        .keyBy(<>)
                        .window(TumblingEventTimeWindows.of(Time.seconds(300)))
                        .allowedLateneess(Time.seconds(900))
                        .sideOutputLateData(lateOutputTag)
                        .aggregate(AggregateFunction, ProcessWindowFunction)
                        .addSink()

我是运行并行度 1，默认 setAutowatermarkInterval 为 200 毫秒。我没有设置 setStreamTimeCharacteristic 因为从 flink 1.12 默认是事件时间。

我看到 StreamWatermarkDebugFilter 的输出正在处理水印，但所有事件都被标记为延迟并在 lateOutputTag 处收集。

2021-05-18 17:14:19,745 INFO                  - ETS: mts: 1621310100000, ts: 1621310582271
2021-05-18 17:14:19,745 INFO                  - ETS: mts: 1621310100000, ts: 1621310582271
2021-05-18 17:14:19,842 INFO  StreamWatermarkDebugFilter         - Op, Watermark: 1621309499999
2021-05-18 17:14:19,944 INFO                  - ETS: mts: 1621309800000, ts: 1621310582275
2021-05-18 17:14:19,944 INFO                  - ETS: mts: 1621309800000, ts: 1621310582275
...
2021-05-18 17:14:20,107 INFO                  - ETS: mts: 1621310380000, ts: 1621310582278
2021-05-18 17:14:20,107 INFO                  - ETS: mts: 1621310380000, ts: 1621310582278
2021-05-18 17:14:20,137 INFO  StreamWatermarkDebugFilter         - Op, Watermark: 1621309779999
2021-05-18 17:14:20,203 INFO                  - ETS: mts: 1621309800000, ts: 1621310582279
...
2021-05-18 17:17:47,839 INFO                  - ETS: mts: 1621310100000, ts: 1621310681159
2021-05-18 17:17:47,848 INFO  StreamWatermarkDebugFilter         - Op, Watermark: 1621310099999
2021-05-18 17:17:47,958 INFO                  - ETS: mts: 1621309800000, ts: 1621310681237
2021-05-18 17:17:47,958 INFO                  - ETS: mts: 1621309800000, ts: 1621310681237
...
2021-05-18 17:22:24,207 INFO                  - ETS: mts: 1621310100000, ts: 1621310703622
2021-05-18 17:22:24,229 INFO  StreamWatermarkDebugFilter         - Op, Watermark: 1621310399999
2021-05-18 17:22:24,315 INFO                  - ETS: mts: 1621309800000, ts: 1621310705177
2021-05-18 17:22:24,315 INFO                  - ETS: mts: 1621309800000, ts: 1621310705177

这个我看过了，不是懒惰的问题

看起来与此有关。有人可以建议我如何进一步调试这个问题以确定可能是什么问题吗？

Answer 1

这是我没有分享的代码部分的问题。我在 assignTimestampsAndWatermarks() 之后做了一个 filter()，所以我不感兴趣的倾斜数据将水印向前推。我将 filter() 移到了 assignTimestampsAndWatermarks 之前，它按预期工作。

flink 1.12.2 所有事件都被删除了

flink 1.12.2 all events getting dropped as late

apache-flink

flink-streaming