flink + Atlas 任务数据血缘调通

atlas flink-bridge分支
参考修改 Flink 源码

	版本
Flink	1.13.5
Atlas	1.2.0

将 atlas 配置文件打进 flink-bridge；atlas 相关的 jar 放进 flink/lib

jar uf flink-bridge-1.2.0.jar atlas-application.properties 

flink-conf.yaml 注册监听

org.apache.flink.configuration.ExecutionOptions 添加配置属性

public static final ConfigOption<List<String>> JOB_LISTENERS =
            ConfigOptions.key("execution.atlas.job-listeners")
                    .stringType()
                    .asList()
                    .noDefaultValue()
                    .withDescription("JobListenerFactories to be registered for the execution.");

一点说明：官方Flink1.12.0 版本之后支持配置execution.job-listeners，因此自己添加了个配置属性execution.atlas.job-listeners 进行区分，
org.apache.flink.configuration.DeploymentOptions

任务提交

flink run -m yarn-cluster -ys 1 -yjm 1024 -ytm 1024 -c com.nufront.bigdata.v2x.test.AtlasTest /opt/v2x-1.0-SNAPSHOT.jar

测试任务

public class AtlasTest {

    public static void main(String[] args) throws Exception {

        //TODO 1.获取执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setParallelism(1);
        env.disableOperatorChaining();

        // TODO kafka消费
        // 配置 kafka 输入流信息
        Properties consumerprops = new Properties();
        consumerprops.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "10.0.2.67:9092");
        consumerprops.put(ConsumerConfig.GROUP_ID_CONFIG, "group1");
        consumerprops.put(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, "false");
        consumerprops.put(ConsumerConfig.AUTO_OFFSET_RESET_CONFIG, "latest");
        consumerprops.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, "org.apache.kafka.common.serialization.StringDeserializer");
        consumerprops.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, "org.apache.kafka.common.serialization.StringDeserializer");
        // 添加 kafka 数据源
        DataStreamSource<String> dataStreamSource = env.addSource(new FlinkKafkaConsumer<>("atlas-source-topic", new SimpleStringSchema(), consumerprops));

        // 配置kafka输入流信息
        Properties producerprops = new Properties();
        producerprops.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "10.0.2.67:9092");
        // 配置证书信息
        dataStreamSource.addSink(new FlinkKafkaProducer<String>("atlas-sink-topic", new KeyedSerializationSchemaWrapper(new SerializationSchema<String>(){
            @Override
            public byte[] serialize(String element) {
                return element.getBytes();
            }
        }), producerprops));
        env.execute("AtlasTest");
    }

}

flink on yarn 日志输出

修改 json 解析方式

org.apache.atlas.utils.AtlasJson#toJson

public static String toJson(Object obj) {
		String ret;

		if (obj instanceof JsonNode && ((JsonNode) obj).isTextual()) {
			ret = ((JsonNode) obj).textValue();
		} else {
			
			// 修改 json 处理方式：fastjson，原来的ObjectMapper.writeValueAsString() 一度卡住不往下执行
			// ret = mapper.writeValueAsString(obj);
			ret = JSONObject.toJSONString(JSONObject.toJSON(obj));
			LOG.info(ret);
		}
		
        return ret;
    }

查看目标 kafka 对应topic

flinksql

Caused by: java.lang.ClassCastException: org.apache.flink.table.runtime.operators.sink.SinkOperator cannot be cast to org.apache.flink.streaming.api.operators.StreamSink

org.apache.atlas.flink.hook.FlinkAtlasHook#addSinkEntities

private void addSinkEntities(List<StreamNode> sinks, AtlasEntity flinkApp, AtlasEntity.AtlasEntityExtInfo entityExtInfo) {
		List<AtlasEntity> outputs = new ArrayList<>();
		for (StreamNode sink : sinks) {

			// StreamSink<?> sinkOperator = (StreamSink<?>) sink.getOperator();
			// SinkFunction<?> sinkFunction = sinkOperator.getUserFunction();
			// List<AtlasEntity> dsEntities = createSinkEntity(sinkFunction, entityExtInfo);
			// outputs.addAll(dsEntities);

			if (sink.getOperator().getClass().getName().equals("org.apache.flink.streaming.api.operators.StreamSink")) {
				StreamSink<?> sinkOperator = (StreamSink<?>) sink.getOperator();
				SinkFunction<?> sinkFunction = sinkOperator.getUserFunction();
				List<AtlasEntity> dsEntities = createSinkEntity(sinkFunction, entityExtInfo);
				outputs.addAll(dsEntities);
			} else if (sink.getOperator().getClass().getName().equals("org.apache.flink.table.runtime.operators.sink.SinkOperator")) {
				SinkOperator sinkOperator = (SinkOperator) sink.getOperator();
				SinkFunction<?> sinkFunction = sinkOperator.getUserFunction();
				List<AtlasEntity> dsEntities = createSinkEntity(sinkFunction, entityExtInfo);
				outputs.addAll(dsEntities);
			}

		}

		flinkApp.setRelationshipAttribute("outputs", AtlasTypeUtil.getAtlasRelatedObjectIds(outputs, RELATIONSHIP_PROCESS_DATASET_OUTPUTS));

		LOG.info("-------------- flinkApp: -------------" + flinkApp);

	}