Amazon Data Firehose를 사용하여 개별 레코드 및 배치 레코드 처리

다음 코드 예제에서는 Firehose를 사용하여 개별 및 배치 레코드를 처리하는 방법을 보여줍니다.

Java

SDK for Java 2.x

참고

GitHub에 더 많은 내용이 있습니다. AWS 코드 예시 리포지토리에서 전체 예시를 찾고 설정 및 실행하는 방법을 배워보세요.

이 예제에서는 개별 및 배치 레코드를 Firehose에 넣습니다.

/**
 * Amazon Firehose Scenario example using Java V2 SDK.
 *
 * Demonstrates individual and batch record processing,
 * and monitoring Firehose delivery stream metrics.
 */
public class FirehoseScenario {

    private static FirehoseClient firehoseClient;
    private static CloudWatchClient cloudWatchClient;

    public static void main(String[] args) {
        final String usage = """
                Usage:
                    <deliveryStreamName>
                Where:
                    deliveryStreamName - The Firehose delivery stream name.
                """;

        if (args.length != 1) {
            System.out.println(usage);
            return;
        }
        String deliveryStreamName = args[0];

        try {
            // Read and parse sample data.
            String jsonContent = readJsonFile("sample_records.json");
            ObjectMapper objectMapper = new ObjectMapper();
            List<Map<String, Object>> sampleData = objectMapper.readValue(jsonContent, new TypeReference<>() {});

            // Process individual records.
            System.out.println("Processing individual records...");
            sampleData.subList(0, 100).forEach(record -> {
                try {
                    putRecord(record, deliveryStreamName);
                } catch (Exception e) {
                    System.err.println("Error processing record: " + e.getMessage());
                }
            });

            // Monitor metrics.
            monitorMetrics(deliveryStreamName);

            // Process batch records.
            System.out.println("Processing batch records...");
            putRecordBatch(sampleData.subList(100, sampleData.size()), 500, deliveryStreamName);
            monitorMetrics(deliveryStreamName);

        } catch (Exception e) {
            System.err.println("Scenario failed: " + e.getMessage());
        } finally {
            closeClients();
        }
    }

    private static FirehoseClient getFirehoseClient() {
        if (firehoseClient == null) {
            firehoseClient = FirehoseClient.create();
        }
        return firehoseClient;
    }

    private static CloudWatchClient getCloudWatchClient() {
        if (cloudWatchClient == null) {
            cloudWatchClient = CloudWatchClient.create();
        }
        return cloudWatchClient;
    }

    /**
     * Puts a record to the specified Amazon Kinesis Data Firehose delivery stream.
     *
     * @param record The record to be put to the delivery stream. The record must be a {@link Map} of String keys and Object values.
     * @param deliveryStreamName The name of the Amazon Kinesis Data Firehose delivery stream to which the record should be put.
     * @throws IllegalArgumentException if the input record or delivery stream name is null or empty.
     * @throws RuntimeException if there is an error putting the record to the delivery stream.
     */
    public static void putRecord(Map<String, Object> record, String deliveryStreamName) {
        if (record == null || deliveryStreamName == null || deliveryStreamName.isEmpty()) {
            throw new IllegalArgumentException("Invalid input: record or delivery stream name cannot be null/empty");
        }
        try {
            String jsonRecord = new ObjectMapper().writeValueAsString(record);
            Record firehoseRecord = Record.builder()
                .data(SdkBytes.fromByteArray(jsonRecord.getBytes(StandardCharsets.UTF_8)))
                .build();

            PutRecordRequest putRecordRequest = PutRecordRequest.builder()
                .deliveryStreamName(deliveryStreamName)
                .record(firehoseRecord)
                .build();

            getFirehoseClient().putRecord(putRecordRequest);
            System.out.println("Record sent: " + jsonRecord);
        } catch (Exception e) {
            throw new RuntimeException("Failed to put record: " + e.getMessage(), e);
        }
    }


    /**
     * Puts a batch of records to an Amazon Kinesis Data Firehose delivery stream.
     *
     * @param records           a list of maps representing the records to be sent
     * @param batchSize         the maximum number of records to include in each batch
     * @param deliveryStreamName the name of the Kinesis Data Firehose delivery stream
     * @throws IllegalArgumentException if the input parameters are invalid (null or empty)
     * @throws RuntimeException         if there is an error putting the record batch
     */
    public static void putRecordBatch(List<Map<String, Object>> records, int batchSize, String deliveryStreamName) {
        if (records == null || records.isEmpty() || deliveryStreamName == null || deliveryStreamName.isEmpty()) {
            throw new IllegalArgumentException("Invalid input: records or delivery stream name cannot be null/empty");
        }
        ObjectMapper objectMapper = new ObjectMapper();

        try {
            for (int i = 0; i < records.size(); i += batchSize) {
                List<Map<String, Object>> batch = records.subList(i, Math.min(i + batchSize, records.size()));

                List<Record> batchRecords = batch.stream().map(record -> {
                    try {
                        String jsonRecord = objectMapper.writeValueAsString(record);
                        return Record.builder()
                            .data(SdkBytes.fromByteArray(jsonRecord.getBytes(StandardCharsets.UTF_8)))
                            .build();
                    } catch (Exception e) {
                        throw new RuntimeException("Error creating Firehose record", e);
                    }
                }).collect(Collectors.toList());

                PutRecordBatchRequest request = PutRecordBatchRequest.builder()
                    .deliveryStreamName(deliveryStreamName)
                    .records(batchRecords)
                    .build();

                PutRecordBatchResponse response = getFirehoseClient().putRecordBatch(request);

                if (response.failedPutCount() > 0) {
                    response.requestResponses().stream()
                        .filter(r -> r.errorCode() != null)
                        .forEach(r -> System.err.println("Failed record: " + r.errorMessage()));
                }
                System.out.println("Batch sent with size: " + batchRecords.size());
            }
        } catch (Exception e) {
            throw new RuntimeException("Failed to put record batch: " + e.getMessage(), e);
        }
    }

    public static void monitorMetrics(String deliveryStreamName) {
        Instant endTime = Instant.now();
        Instant startTime = endTime.minusSeconds(600);

        List<String> metrics = List.of("IncomingBytes", "IncomingRecords", "FailedPutCount");
        metrics.forEach(metric -> monitorMetric(metric, startTime, endTime, deliveryStreamName));
    }

    private static void monitorMetric(String metricName, Instant startTime, Instant endTime, String deliveryStreamName) {
        try {
            GetMetricStatisticsRequest request = GetMetricStatisticsRequest.builder()
                .namespace("AWS/Firehose")
                .metricName(metricName)
                .dimensions(Dimension.builder().name("DeliveryStreamName").value(deliveryStreamName).build())
                .startTime(startTime)
                .endTime(endTime)
                .period(60)
                .statistics(Statistic.SUM)
                .build();

            GetMetricStatisticsResponse response = getCloudWatchClient().getMetricStatistics(request);
            double totalSum = response.datapoints().stream().mapToDouble(Datapoint::sum).sum();
            System.out.println(metricName + ": " + totalSum);
        } catch (Exception e) {
            System.err.println("Failed to monitor metric " + metricName + ": " + e.getMessage());
        }
    }

    public static String readJsonFile(String fileName) throws IOException {
        try (InputStream inputStream = FirehoseScenario.class.getResourceAsStream("/" + fileName);
             Scanner scanner = new Scanner(inputStream, StandardCharsets.UTF_8)) {
            return scanner.useDelimiter("\\\\A").next();
        } catch (Exception e) {
            throw new RuntimeException("Error reading file: " + fileName, e);
        }
    }

    private static void closeClients() {
        try {
            if (firehoseClient != null) firehoseClient.close();
            if (cloudWatchClient != null) cloudWatchClient.close();
        } catch (Exception e) {
            System.err.println("Error closing clients: " + e.getMessage());
        }
    }
}

• API 세부 정보는 AWS SDK for Java 2.x API 참조의 다음 주제를 참조하십시오.

  • PutRecord

  • PutRecordBatch

Python

SDK for Python (Boto3)

참고

GitHub에 더 많은 내용이 있습니다. AWS 코드 예시 리포지토리에서 전체 예시를 찾고 설정 및 실행하는 방법을 배워보세요.

이 스크립트는 개별 레코드 및 배치 레코드를 Firehose에 넣습니다.

import json
import logging
import random
from datetime import datetime, timedelta

import backoff
import boto3

from config import get_config


def load_sample_data(path: str) -> dict:
    """
    Load sample data from a JSON file.

    Args:
        path (str): The file path to the JSON file containing sample data.

    Returns:
        dict: The loaded sample data as a dictionary.
    """
    with open(path, "r") as f:
        return json.load(f)


# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)


class FirehoseClient:
    """
    AWS Firehose client to send records and monitor metrics.

    Attributes:
        config (object): Configuration object with delivery stream name and region.
        delivery_stream_name (str): Name of the Firehose delivery stream.
        region (str): AWS region for Firehose and CloudWatch clients.
        firehose (boto3.client): Boto3 Firehose client.
        cloudwatch (boto3.client): Boto3 CloudWatch client.
    """

    def __init__(self, config):
        """
        Initialize the FirehoseClient.

        Args:
            config (object): Configuration object with delivery stream name and region.
        """
        self.config = config
        self.delivery_stream_name = config.delivery_stream_name
        self.region = config.region
        self.firehose = boto3.client("firehose", region_name=self.region)
        self.cloudwatch = boto3.client("cloudwatch", region_name=self.region)


    @backoff.on_exception(
        backoff.expo, Exception, max_tries=5, jitter=backoff.full_jitter
    )
    def put_record(self, record: dict):
        """
        Put individual records to Firehose with backoff and retry.

        Args:
            record (dict): The data record to be sent to Firehose.

        This method attempts to send an individual record to the Firehose delivery stream.
        It retries with exponential backoff in case of exceptions.
        """
        try:
            entry = self._create_record_entry(record)
            response = self.firehose.put_record(
                DeliveryStreamName=self.delivery_stream_name, Record=entry
            )
            self._log_response(response, entry)
        except Exception:
            logger.info(f"Fail record: {record}.")
            raise


    @backoff.on_exception(
        backoff.expo, Exception, max_tries=5, jitter=backoff.full_jitter
    )
    def put_record_batch(self, data: list, batch_size: int = 500):
        """
        Put records in batches to Firehose with backoff and retry.

        Args:
            data (list): List of data records to be sent to Firehose.
            batch_size (int): Number of records to send in each batch. Default is 500.

        This method attempts to send records in batches to the Firehose delivery stream.
        It retries with exponential backoff in case of exceptions.
        """
        for i in range(0, len(data), batch_size):
            batch = data[i : i + batch_size]
            record_dicts = [{"Data": json.dumps(record)} for record in batch]
            try:
                response = self.firehose.put_record_batch(
                    DeliveryStreamName=self.delivery_stream_name, Records=record_dicts
                )
                self._log_batch_response(response, len(batch))
            except Exception as e:
                logger.info(f"Failed to send batch of {len(batch)} records. Error: {e}")


    def get_metric_statistics(
        self,
        metric_name: str,
        start_time: datetime,
        end_time: datetime,
        period: int,
        statistics: list = ["Sum"],
    ) -> list:
        """
        Retrieve metric statistics from CloudWatch.

        Args:
            metric_name (str): The name of the metric.
            start_time (datetime): The start time for the metric statistics.
            end_time (datetime): The end time for the metric statistics.
            period (int): The granularity, in seconds, of the returned data points.
            statistics (list): A list of statistics to retrieve. Default is ['Sum'].

        Returns:
            list: List of datapoints containing the metric statistics.
        """
        response = self.cloudwatch.get_metric_statistics(
            Namespace="AWS/Firehose",
            MetricName=metric_name,
            Dimensions=[
                {"Name": "DeliveryStreamName", "Value": self.delivery_stream_name},
            ],
            StartTime=start_time,
            EndTime=end_time,
            Period=period,
            Statistics=statistics,
        )
        return response["Datapoints"]

    def monitor_metrics(self):
        """
        Monitor Firehose metrics for the last 5 minutes.

        This method retrieves and logs the 'IncomingBytes', 'IncomingRecords', and 'FailedPutCount' metrics
        from CloudWatch for the last 5 minutes.
        """
        end_time = datetime.utcnow()
        start_time = end_time - timedelta(minutes=10)
        period = int((end_time - start_time).total_seconds())

        metrics = {
            "IncomingBytes": self.get_metric_statistics(
                "IncomingBytes", start_time, end_time, period
            ),
            "IncomingRecords": self.get_metric_statistics(
                "IncomingRecords", start_time, end_time, period
            ),
            "FailedPutCount": self.get_metric_statistics(
                "FailedPutCount", start_time, end_time, period
            ),
        }

        for metric, datapoints in metrics.items():
            if datapoints:
                total_sum = sum(datapoint["Sum"] for datapoint in datapoints)
                if metric == "IncomingBytes":
                    logger.info(
                        f"{metric}: {round(total_sum)} ({total_sum / (1024 * 1024):.2f} MB)"
                    )
                else:
                    logger.info(f"{metric}: {round(total_sum)}")
            else:
                logger.info(f"No data found for {metric} over the last 5 minutes")


    def _create_record_entry(self, record: dict) -> dict:
        """
        Create a record entry for Firehose.

        Args:
            record (dict): The data record to be sent.

        Returns:
            dict: The record entry formatted for Firehose.

        Raises:
            Exception: If a simulated network error occurs.
        """
        if random.random() < 0.2:
            raise Exception("Simulated network error")
        elif random.random() < 0.1:
            return {"Data": '{"malformed": "data"'}
        else:
            return {"Data": json.dumps(record)}

    def _log_response(self, response: dict, entry: dict):
        """
        Log the response from Firehose.

        Args:
            response (dict): The response from the Firehose put_record API call.
            entry (dict): The record entry that was sent.
        """
        if response["ResponseMetadata"]["HTTPStatusCode"] == 200:
            logger.info(f"Sent record: {entry}")
        else:
            logger.info(f"Fail record: {entry}")

    def _log_batch_response(self, response: dict, batch_size: int):
        """
        Log the batch response from Firehose.

        Args:
            response (dict): The response from the Firehose put_record_batch API call.
            batch_size (int): The number of records in the batch.
        """
        if response.get("FailedPutCount", 0) > 0:
            logger.info(
                f'Failed to send {response["FailedPutCount"]} records in batch of {batch_size}'
            )
        else:
            logger.info(f"Successfully sent batch of {batch_size} records")


if __name__ == "__main__":
    config = get_config()
    data = load_sample_data(config.sample_data_file)
    client = FirehoseClient(config)

    # Process the first 100 sample network records
    for record in data[:100]:
        try:
            client.put_record(record)
        except Exception as e:
            logger.info(f"Put record failed after retries and backoff: {e}")
    client.monitor_metrics()

    # Process remaining records using the batch method
    try:
        client.put_record_batch(data[100:])
    except Exception as e:
        logger.info(f"Put record batch failed after retries and backoff: {e}")
    client.monitor_metrics()

이 파일에는 위 스크립트에 대한 구성이 포함되어 있습니다.

class Config:
    def __init__(self):
        self.delivery_stream_name = "ENTER YOUR DELIVERY STREAM NAME HERE"
        self.region = "us-east-1"
        self.sample_data_file = (
            "../../../../../scenarios/features/firehose/resources/sample_records.json"
        )


def get_config():
    return Config()

• API 세부 정보는 AWS SDK for Python (Boto3) API 참조의 다음 주제를 참조하세요.

  • PutRecord

  • PutRecordBatch

AWS SDK 개발자 가이드 및 코드 예제의 전체 목록은 섹션을 참조하세요AWS SDK와 함께 Firehose 사용. 이 주제에는 시작하기에 대한 정보와 이전 SDK 버전에 대한 세부 정보도 포함되어 있습니다.