Pub/Sub to MongoDB with Python UDF 템플릿

Pub/Sub to MongoDB with Python UDF 템플릿은 Pub/Sub 구독에서 JSON 인코딩 메시지를 읽고 이를 MongoDB에 문서로 쓰는 스트리밍 파이프라인입니다. 필요한 경우 이 파이프라인은 Python 사용자 정의 함수(UDF)를 사용하여 포함할 수 있는 추가 변환을 지원합니다.

레코드를 처리하는 중에 오류가 발생하면 템플릿은 입력 메시지와 함께 BigQuery 테이블에 레코드를 씁니다. 예를 들어 스키마 불일치나 잘못된 형식의 JSON으로 인해 또는 변환 실행 중에 오류가 발생할 수 있습니다. deadletterTable 매개변수에 테이블 이름을 지정합니다. 테이블이 없으면 파이프라인에서 자동으로 만듭니다.

파이프라인 요구사항

  • Pub/Sub 구독이 있어야 하며 메시지가 유효한 JSON 형식으로 인코딩되어야 합니다.
  • MongoDB 클러스터가 있어야 하며 Dataflow 작업자 머신에서 액세스할 수 있어야 합니다.

템플릿 매개변수

매개변수 설명
inputSubscription Pub/Sub 구독의 이름입니다. 예를 들면 projects/my-project-id/subscriptions/my-subscription-id입니다.
mongoDBUri 쉼표로 구분된 MongoDB 서버 목록입니다. 예를 들면 192.285.234.12:27017,192.287.123.11:27017입니다.
database 컬렉션을 저장하기 위한 MongoDB의 데이터베이스입니다. 예를 들면 my-db입니다.
collection MongoDB 데이터베이스 내부의 컬렉션 이름입니다. 예를 들면 my-collection입니다.
deadletterTable 오류(스키마 불일치, 잘못된 형식의 JSON 등)로 인해 메시지를 저장하는 BigQuery 테이블입니다. 예를 들면 project-id:dataset-name.table-name입니다.
pythonExternalTextTransformGcsPath (선택사항) 사용할 사용자 정의 함수(UDF)를 정의하는 Python 코드 파일의 Cloud Storage URI입니다. 예를 들면 gs://my-bucket/my-udfs/my_file.py입니다.
pythonExternalTextTransformFunctionName (선택사항) 사용할 Python 사용자 정의 함수(UDF)의 이름입니다.
batchSize 선택사항: MongoDB에 문서를 일괄 삽입하기 위해 사용되는 배치 크기입니다. 기본값: 1000
batchSizeBytes 선택사항: 배치 크기(바이트)입니다. 기본값: 5242880
maxConnectionIdleTime 선택사항: 연결이 시간 초과되기 전까지 허용되는 최대 유휴 시간(초)입니다. 기본값: 60000
sslEnabled 선택사항: MongoDB 연결에 SSL이 사용 설정되었는지 여부를 나타내는 부울 값입니다. 기본값: true
ignoreSSLCertificate 선택사항: SSL 인증서를 무시해야 하는지 여부를 나타내는 부울 값입니다. 기본값: true
withOrdered 선택사항: MongoDB에 순서대로 대량 삽입할 수 있게 해주는 부울 값입니다. 기본값: true
withSSLInvalidHostNameAllowed 선택사항: SSL 연결에 잘못된 호스트 이름이 허용되었는지 여부를 나타내는 부울 값입니다. 기본값: true

사용자 정의 함수

선택적으로 사용자 정의 함수(UDF)를 작성하여 이 템플릿을 확장할 수 있습니다. 템플릿이 각 입력 요소에 대해 UDF를 호출합니다. 요소 페이로드는 JSON 문자열로 직렬화됩니다. 자세한 내용은 Dataflow 템플릿에 대한 사용자 정의 함수 만들기를 참조하세요.

함수 사양

UDF의 사양은 다음과 같습니다.

  • 입력: 입력 CSV 파일의 한 줄입니다.
  • 출력: MongoDB에 삽입할 문자열화된 JSON 문서입니다.

템플릿 실행

콘솔gcloudAPI
  1. Dataflow 템플릿에서 작업 만들기 페이지로 이동합니다.
  2. 템플릿에서 작업 만들기로 이동
  3. 작업 이름 필드에 고유한 작업 이름을 입력합니다.
  4. (선택사항): 리전 엔드포인트의 드롭다운 메뉴에서 값을 선택합니다. 기본 리전은 us-central1입니다.

    Dataflow 작업을 실행할 수 있는 리전 목록은 Dataflow 위치를 참조하세요.

  5. Dataflow 템플릿 드롭다운 메뉴에서 the Pub/Sub to MongoDB with Python UDFs template을 선택합니다.
  6. 제공된 매개변수 필드에 매개변수 값을 입력합니다.
  7. 작업 실행을 클릭합니다.

셸 또는 터미널에서 템플릿을 실행합니다.

gcloud dataflow flex-template run JOB_NAME \
    --project=PROJECT_ID \
    --region=REGION_NAME \
    --template-file-gcs-location=gs://dataflow-templates-REGION_NAME/VERSION/flex/Cloud_PubSub_to_MongoDB_Xlang \
    --parameters \
inputSubscription=INPUT_SUBSCRIPTION,\
mongoDBUri=MONGODB_URI,\
database=DATABASE,
collection=COLLECTION,
deadletterTable=UNPROCESSED_TABLE
  

다음을 바꿉니다.

  • PROJECT_ID: Dataflow 작업을 실행하려는 Google Cloud 프로젝트 ID
  • REGION_NAME: Dataflow 작업을 배포할 리전(예: us-central1)
  • JOB_NAME: 선택한 고유한 작업 이름
  • VERSION: 사용할 템플릿 버전

    다음 값을 사용할 수 있습니다.

  • INPUT_SUBSCRIPTION: Pub/Sub 구독(예: projects/my-project-id/subscriptions/my-subscription-id)
  • MONGODB_URI: MongoDB 서버 주소(예: 192.285.234.12:27017,192.287.123.11:27017)
  • DATABASE: MongoDB 데이터베이스의 이름(예: users)
  • COLLECTION: MongoDB 컬렉션의 이름(예: profiles)
  • UNPROCESSED_TABLE: BigQuery 테이블의 이름(예: your-project:your-dataset.your-table-name)

REST API를 사용하여 템플릿을 실행하려면 HTTP POST 요청을 전송합니다. API 및 승인 범위에 대한 자세한 내용은 projects.templates.launch를 참조하세요.

POST https://dataflow.googleapis.com/v1b3/projects/PROJECT_ID/locations/LOCATION/flexTemplates:launch
{
   "launch_parameter": {
      "jobName": "JOB_NAME",
      "parameters": {
          "inputSubscription": "INPUT_SUBSCRIPTION",
          "mongoDBUri": "MONGODB_URI",
          "database": "DATABASE",
          "collection": "COLLECTION",
          "deadletterTable": "UNPROCESSED_TABLE"
      },
      "containerSpecGcsPath": "gs://dataflow-templates-LOCATION/VERSION/flex/Cloud_PubSub_to_MongoDB_Xlang",
   }
}
  

다음을 바꿉니다.

  • PROJECT_ID: Dataflow 작업을 실행하려는 Google Cloud 프로젝트 ID
  • LOCATION: Dataflow 작업을 배포할 리전(예: us-central1)
  • JOB_NAME: 선택한 고유한 작업 이름
  • VERSION: 사용할 템플릿 버전

    다음 값을 사용할 수 있습니다.

  • INPUT_SUBSCRIPTION: Pub/Sub 구독(예: projects/my-project-id/subscriptions/my-subscription-id)
  • MONGODB_URI: MongoDB 서버 주소(예: 192.285.234.12:27017,192.287.123.11:27017)
  • DATABASE: MongoDB 데이터베이스의 이름(예: users)
  • COLLECTION: MongoDB 컬렉션의 이름(예: profiles)
  • UNPROCESSED_TABLE: BigQuery 테이블의 이름(예: your-project:your-dataset.your-table-name)
Java
/*
 * Copyright (C) 2019 Google LLC
 *
 * Licensed under the Apache License, Version 2.0 (the "License"); you may not
 * use this file except in compliance with the License. You may obtain a copy of
 * the License at
 *
 *   http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
 * License for the specific language governing permissions and limitations under
 * the License.
 */
package com.google.cloud.teleport.v2.templates;

import com.google.auto.value.AutoValue;
import com.google.cloud.teleport.metadata.MultiTemplate;
import com.google.cloud.teleport.metadata.Template;
import com.google.cloud.teleport.metadata.TemplateCategory;
import com.google.cloud.teleport.metadata.TemplateParameter;
import com.google.cloud.teleport.v2.coders.FailsafeElementCoder;
import com.google.cloud.teleport.v2.common.UncaughtExceptionLogger;
import com.google.cloud.teleport.v2.templates.PubSubToMongoDB.Options;
import com.google.cloud.teleport.v2.transforms.ErrorConverters;
import com.google.cloud.teleport.v2.transforms.JavascriptTextTransformer;
import com.google.cloud.teleport.v2.transforms.PythonExternalTextTransformer;
import com.google.cloud.teleport.v2.utils.SchemaUtils;
import com.google.cloud.teleport.v2.values.FailsafeElement;
import com.google.common.base.Strings;
import com.google.gson.Gson;
import com.google.gson.JsonObject;
import com.google.gson.JsonSyntaxException;
import java.nio.charset.StandardCharsets;
import javax.annotation.Nullable;
import org.apache.beam.sdk.Pipeline;
import org.apache.beam.sdk.PipelineResult;
import org.apache.beam.sdk.coders.CoderRegistry;
import org.apache.beam.sdk.coders.StringUtf8Coder;
import org.apache.beam.sdk.io.gcp.pubsub.PubsubIO;
import org.apache.beam.sdk.io.gcp.pubsub.PubsubMessage;
import org.apache.beam.sdk.io.gcp.pubsub.PubsubMessageWithAttributesCoder;
import org.apache.beam.sdk.io.mongodb.MongoDbIO;
import org.apache.beam.sdk.metrics.Counter;
import org.apache.beam.sdk.metrics.Metrics;
import org.apache.beam.sdk.options.Default;
import org.apache.beam.sdk.options.PipelineOptions;
import org.apache.beam.sdk.options.PipelineOptionsFactory;
import org.apache.beam.sdk.options.Validation;
import org.apache.beam.sdk.transforms.DoFn;
import org.apache.beam.sdk.transforms.MapElements;
import org.apache.beam.sdk.transforms.PTransform;
import org.apache.beam.sdk.transforms.ParDo;
import org.apache.beam.sdk.values.PCollection;
import org.apache.beam.sdk.values.PCollectionTuple;
import org.apache.beam.sdk.values.Row;
import org.apache.beam.sdk.values.TupleTag;
import org.apache.beam.sdk.values.TupleTagList;
import org.apache.beam.sdk.values.TypeDescriptors;
import org.apache.beam.vendor.guava.v32_1_2_jre.com.google.common.base.Throwables;
import org.bson.Document;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

/**
 * The {@link PubSubToMongoDB} pipeline is a streaming pipeline which ingests data in JSON format
 * from PubSub, applies a Javascript UDF if provided and inserts resulting records as Bson Document
 * in MongoDB. If the element fails to be processed then it is written to a deadletter table in
 * BigQuery.
 *
 * <p><b>Pipeline Requirements</b>
 *
 * <ul>
 *   <li>The PubSub topic and subscriptions exist
 *   <li>The MongoDB is up and running
 * </ul>
 *
 * <p>Check out <a
 * href="https://github.com/GoogleCloudPlatform/DataflowTemplates/blob/main/v2/pubsub-to-mongodb/README_Cloud_PubSub_to_MongoDB.md">README</a>
 * for instructions on how to use or modify this template.
 */
@MultiTemplate({
  @Template(
      name = "Cloud_PubSub_to_MongoDB",
      category = TemplateCategory.STREAMING,
      displayName = "Pub/Sub to MongoDB",
      description =
          "The Pub/Sub to MongoDB template is a streaming pipeline that reads JSON-encoded messages from a Pub/Sub subscription and writes them to MongoDB as documents. "
              + "If required, this pipeline supports additional transforms that can be included using a JavaScript user-defined function (UDF). "
              + "Any errors occurred due to schema mismatch, malformed JSON, or while executing transforms are recorded in a BigQuery table for unprocessed messages along with input message. "
              + "If a table for unprocessed records does not exist prior to execution, the pipeline automatically creates this table.",
      skipOptions = {
        "pythonExternalTextTransformGcsPath",
        "pythonExternalTextTransformFunctionName"
      },
      optionsClass = Options.class,
      flexContainerName = "pubsub-to-mongodb",
      documentation =
          "https://cloud.google.com/dataflow/docs/guides/templates/provided/pubsub-to-mongodb",
      contactInformation = "https://cloud.google.com/support",
      preview = true,
      requirements = {
        "The Pub/Sub Subscription must exist and the messages must be encoded in a valid JSON format.",
        "The MongoDB cluster must exist and should be accessible from the Dataflow worker machines."
      },
      streaming = true,
      supportsAtLeastOnce = true),
  @Template(
      name = "Cloud_PubSub_to_MongoDB_Xlang",
      category = TemplateCategory.STREAMING,
      type = Template.TemplateType.XLANG,
      displayName = "Pub/Sub to MongoDB with Python UDFs",
      description =
          "The Pub/Sub to MongoDB template is a streaming pipeline that reads JSON-encoded messages from a Pub/Sub subscription and writes them to MongoDB as documents. "
              + "If required, this pipeline supports additional transforms that can be included using a Python user-defined function (UDF). "
              + "Any errors occurred due to schema mismatch, malformed JSON, or while executing transforms are recorded in a BigQuery table for unprocessed messages along with input message. "
              + "If a table for unprocessed records does not exist prior to execution, the pipeline automatically creates this table.",
      skipOptions = {
        "javascriptTextTransformGcsPath",
        "javascriptTextTransformFunctionName",
        "javascriptTextTransformReloadIntervalMinutes"
      },
      optionsClass = Options.class,
      flexContainerName = "pubsub-to-mongodb-xlang",
      documentation =
          "https://cloud.google.com/dataflow/docs/guides/templates/provided/pubsub-to-mongodb",
      contactInformation = "https://cloud.google.com/support",
      preview = true,
      requirements = {
        "The Pub/Sub Subscription must exist and the messages must be encoded in a valid JSON format.",
        "The MongoDB cluster must exist and should be accessible from the Dataflow worker machines."
      },
      streaming = true,
      supportsAtLeastOnce = true)
})
public class PubSubToMongoDB {
  /**
   * Options supported by {@link PubSubToMongoDB}
   *
   * <p>Inherits standard configuration options.
   */

  /** The tag for the main output of the json transformation. */
  public static final TupleTag<FailsafeElement<PubsubMessage, String>> TRANSFORM_OUT =
      new TupleTag<FailsafeElement<PubsubMessage, String>>() {};

  /** The tag for the dead-letter output of the json to table row transform. */
  public static final TupleTag<FailsafeElement<PubsubMessage, String>> TRANSFORM_DEADLETTER_OUT =
      new TupleTag<FailsafeElement<PubsubMessage, String>>() {};

  /** Pubsub message/string coder for pipeline. */
  public static final FailsafeElementCoder<PubsubMessage, String> CODER =
      FailsafeElementCoder.of(PubsubMessageWithAttributesCoder.of(), StringUtf8Coder.of());

  /** String/String Coder for FailsafeElement. */
  public static final FailsafeElementCoder<String, String> FAILSAFE_ELEMENT_CODER =
      FailsafeElementCoder.of(StringUtf8Coder.of(), StringUtf8Coder.of());

  /** The log to output status messages to. */
  private static final Logger LOG = LoggerFactory.getLogger(PubSubToMongoDB.class);

  /**
   * The {@link Options} class provides the custom execution options passed by the executor at the
   * command-line.
   *
   * <p>Inherits standard configuration options, options from {@link
   * PythonExternalTextTransformer.PythonExternalTextTransformerOptions}.
   */
  public interface Options
      extends PythonExternalTextTransformer.PythonExternalTextTransformerOptions, PipelineOptions {
    @TemplateParameter.PubsubSubscription(
        order = 1,
        groupName = "Source",
        description = "Pub/Sub input subscription",
        helpText = "Name of the Pub/Sub subscription.",
        example = "projects/your-project-id/subscriptions/your-subscription-name")
    @Validation.Required
    String getInputSubscription();

    void setInputSubscription(String inputSubscription);

    @TemplateParameter.Text(
        order = 2,
        groupName = "Target",
        description = "MongoDB Connection URI",
        helpText = "Comma separated list of MongoDB servers.",
        example = "host1:port,host2:port,host3:port")
    @Validation.Required
    String getMongoDBUri();

    void setMongoDBUri(String mongoDBUri);

    @TemplateParameter.Text(
        order = 3,
        groupName = "Target",
        description = "MongoDB Database",
        helpText = "Database in MongoDB to store the collection.",
        example = "my-db")
    @Validation.Required
    String getDatabase();

    void setDatabase(String database);

    @TemplateParameter.Text(
        order = 4,
        groupName = "Target",
        description = "MongoDB collection",
        helpText = "Name of the collection in the MongoDB database.",
        example = "my-collection")
    @Validation.Required
    String getCollection();

    void setCollection(String collection);

    @TemplateParameter.BigQueryTable(
        order = 5,
        description = "The dead-letter table name to output failed messages to BigQuery",
        helpText =
            "The BigQuery table that stores messages caused by failures, such as mismatched schema, malformed JSON, and so on.",
        example = "your-project-id:your-dataset.your-table-name")
    @Validation.Required
    String getDeadletterTable();

    void setDeadletterTable(String deadletterTable);

    @TemplateParameter.Long(
        order = 6,
        optional = true,
        description = "Batch Size",
        helpText = "Batch size used for batch insertion of documents into MongoDB.")
    @Default.Long(1000)
    Long getBatchSize();

    void setBatchSize(Long batchSize);

    @TemplateParameter.Long(
        order = 7,
        optional = true,
        description = "Batch Size in Bytes",
        helpText = "Batch size in bytes.")
    @Default.Long(5242880)
    Long getBatchSizeBytes();

    void setBatchSizeBytes(Long batchSizeBytes);

    @TemplateParameter.Integer(
        order = 8,
        optional = true,
        description = "Max Connection idle time",
        helpText = "Maximum idle time allowed in seconds before connection timeout occurs.")
    @Default.Integer(60000)
    int getMaxConnectionIdleTime();

    void setMaxConnectionIdleTime(int maxConnectionIdleTime);

    @TemplateParameter.Boolean(
        order = 9,
        optional = true,
        description = "SSL Enabled",
        helpText = "Boolean value indicating whether the connection to MongoDB is SSL enabled.")
    @Default.Boolean(true)
    Boolean getSslEnabled();

    void setSslEnabled(Boolean sslEnabled);

    @TemplateParameter.Boolean(
        order = 10,
        optional = true,
        description = "Ignore SSL Certificate",
        helpText = "Boolean value indicating whether to ignore the SSL certificate.")
    @Default.Boolean(true)
    Boolean getIgnoreSSLCertificate();

    void setIgnoreSSLCertificate(Boolean ignoreSSLCertificate);

    @TemplateParameter.Boolean(
        order = 11,
        optional = true,
        description = "withOrdered",
        helpText = "Boolean value enabling ordered bulk insertions into MongoDB.")
    @Default.Boolean(true)
    Boolean getWithOrdered();

    void setWithOrdered(Boolean withOrdered);

    @TemplateParameter.Boolean(
        order = 12,
        optional = true,
        description = "withSSLInvalidHostNameAllowed",
        helpText =
            "Boolean value indicating whether an invalid hostname is allowed for the SSL connection.")
    @Default.Boolean(true)
    Boolean getWithSSLInvalidHostNameAllowed();

    void setWithSSLInvalidHostNameAllowed(Boolean withSSLInvalidHostNameAllowed);
  }

  /** DoFn that will parse the given string elements as Bson Documents. */
  private static class ParseAsDocumentsFn extends DoFn<String, Document> {

    @ProcessElement
    public void processElement(ProcessContext context) {
      context.output(Document.parse(context.element()));
    }
  }

  /**
   * Main entry point for executing the pipeline.
   *
   * @param args The command-line arguments to the pipeline.
   */
  public static void main(String[] args) {
    UncaughtExceptionLogger.register();

    // Parse the user options passed from the command-line.
    Options options = PipelineOptionsFactory.fromArgs(args).withValidation().as(Options.class);
    run(options);
  }

  /**
   * Runs the pipeline with the supplied options.
   *
   * @param options The execution parameters to the pipeline.
   * @return The result of the pipeline execution.
   */
  public static PipelineResult run(Options options) {

    // Create the pipeline
    Pipeline pipeline = Pipeline.create(options);

    // Register the coders for pipeline
    CoderRegistry coderRegistry = pipeline.getCoderRegistry();

    coderRegistry.registerCoderForType(
        FAILSAFE_ELEMENT_CODER.getEncodedTypeDescriptor(), FAILSAFE_ELEMENT_CODER);

    coderRegistry.registerCoderForType(CODER.getEncodedTypeDescriptor(), CODER);

    boolean usePythonUdf = !Strings.isNullOrEmpty(options.getPythonExternalTextTransformGcsPath());
    boolean useJavascriptUdf = !Strings.isNullOrEmpty(options.getJavascriptTextTransformGcsPath());

    if (usePythonUdf && useJavascriptUdf) {
      throw new IllegalArgumentException(
          "Either javascript or Python gcs path must be provided, but not both.");
    }
    /*
     * Steps: 1) Read PubSubMessage with attributes from input PubSub subscription.
     *        2) Apply Javascript or Python UDF if provided.
     *        3) Write to MongoDB
     *
     */

    LOG.info("Reading from subscription: " + options.getInputSubscription());

    PCollection<PubsubMessage> readMessagesFromPubsub =
        pipeline
            /*
             * Step #1: Read from a PubSub subscription.
             */
            .apply(
            "Read PubSub Subscription",
            PubsubIO.readMessagesWithAttributes().fromSubscription(options.getInputSubscription()));

    PCollectionTuple convertedPubsubMessages;
    if (usePythonUdf) {
      convertedPubsubMessages =
          readMessagesFromPubsub
              /*
               * Step #2: Apply Python Transform and transform, if provided and transform
               *          the PubsubMessages into Json documents.
               */
              .apply(
              "Apply Python UDF",
              PubSubMessageToJsonDocument.newBuilder()
                  .setPythonExternalTextTransformFunctionName(
                      options.getPythonExternalTextTransformFunctionName())
                  .setPythonExternalTextTransformGcsPath(
                      options.getPythonExternalTextTransformGcsPath())
                  .build());
    } else {
      convertedPubsubMessages =
          readMessagesFromPubsub
              /*
               * Step #2: Apply Javascript Transform and transform, if provided and transform
               *          the PubsubMessages into Json documents.
               */
              .apply(
              "Apply Javascript UDF",
              PubSubMessageToJsonDocument.newBuilder()
                  .setJavascriptTextTransformFunctionName(
                      options.getJavascriptTextTransformFunctionName())
                  .setJavascriptTextTransformGcsPath(options.getJavascriptTextTransformGcsPath())
                  .setJavascriptTextTransformReloadIntervalMinutes(
                      options.getJavascriptTextTransformReloadIntervalMinutes())
                  .build());
    }

    /*
     * Step #3a: Write Json documents into MongoDB using {@link MongoDbIO.write}.
     */
    convertedPubsubMessages
        .get(TRANSFORM_OUT)
        .apply(
            "Get Json Documents",
            MapElements.into(TypeDescriptors.strings()).via(FailsafeElement::getPayload))
        .apply("Parse as BSON Document", ParDo.of(new ParseAsDocumentsFn()))
        .apply(
            "Put to MongoDB",
            MongoDbIO.write()
                .withBatchSize(options.getBatchSize())
                .withUri(prefixMongoDb(options.getMongoDBUri()))
                .withDatabase(options.getDatabase())
                .withCollection(options.getCollection())
                .withIgnoreSSLCertificate(options.getIgnoreSSLCertificate())
                .withMaxConnectionIdleTime(options.getMaxConnectionIdleTime())
                .withOrdered(options.getWithOrdered())
                .withSSLEnabled(options.getSslEnabled())
                .withSSLInvalidHostNameAllowed(options.getWithSSLInvalidHostNameAllowed()));

    /*
     * Step 3b: Write elements that failed processing to deadletter table via {@link BigQueryIO}.
     */
    convertedPubsubMessages
        .get(TRANSFORM_DEADLETTER_OUT)
        .apply(
            "Write Transform Failures To BigQuery",
            ErrorConverters.WritePubsubMessageErrors.newBuilder()
                .setErrorRecordsTable(options.getDeadletterTable())
                .setErrorRecordsTableSchema(SchemaUtils.DEADLETTER_SCHEMA)
                .build());

    // Execute the pipeline and return the result.
    return pipeline.run();
  }

  /** Add the MongoDB protocol prefix only if the given uri doesn't have it. */
  private static String prefixMongoDb(String mongoDBUri) {
    if (mongoDBUri.startsWith("mongodb://") || mongoDBUri.startsWith("mongodb+srv://")) {
      return mongoDBUri;
    }
    return String.format("mongodb://%s", mongoDBUri);
  }

  /**
   * The {@link PubSubMessageToJsonDocument} class is a {@link PTransform} which transforms incoming
   * {@link PubsubMessage} objects into JSON objects for insertion into MongoDB while applying an
   * optional UDF to the input. The executions of the UDF and transformation to Json objects is done
   * in a fail-safe way by wrapping the element with it's original payload inside the {@link
   * FailsafeElement} class. The {@link PubSubMessageToJsonDocument} transform will output a {@link
   * PCollectionTuple} which contains all output and dead-letter {@link PCollection}.
   *
   * <p>The {@link PCollectionTuple} output will contain the following {@link PCollection}:
   *
   * <ul>
   *   <li>{@link PubSubToMongoDB#TRANSFORM_OUT} - Contains all records successfully converted to
   *       JSON objects.
   *   <li>{@link PubSubToMongoDB#TRANSFORM_DEADLETTER_OUT} - Contains all {@link FailsafeElement}
   *       records which couldn't be converted to table rows.
   * </ul>
   */
  @AutoValue
  public abstract static class PubSubMessageToJsonDocument
      extends PTransform<PCollection<PubsubMessage>, PCollectionTuple> {

    public static Builder newBuilder() {
      return new AutoValue_PubSubToMongoDB_PubSubMessageToJsonDocument.Builder();
    }

    @Nullable
    public abstract String javascriptTextTransformGcsPath();

    @Nullable
    public abstract String javascriptTextTransformFunctionName();

    @Nullable
    public abstract String pythonExternalTextTransformGcsPath();

    @Nullable
    public abstract String pythonExternalTextTransformFunctionName();

    @Nullable
    public abstract Integer javascriptTextTransformReloadIntervalMinutes();

    @Override
    public PCollectionTuple expand(PCollection<PubsubMessage> input) {

      // Check for python UDF first. If the pipeline is not using Python UDF, proceed as normal.
      if (pythonExternalTextTransformGcsPath() != null) {
        PCollection<Row> failsafeElements =
            input
                // Map the incoming messages into FailsafeElements, so we can recover from failures
                // across multiple transforms.
                .apply(
                    "MapToRecord",
                    PythonExternalTextTransformer.FailsafeRowPythonExternalUdf
                        .pubSubMappingFunction())
                .setRowSchema(
                    PythonExternalTextTransformer.FailsafeRowPythonExternalUdf.ROW_SCHEMA);
        return failsafeElements
            .apply(
                "InvokeUDF",
                PythonExternalTextTransformer.FailsafePythonExternalUdf.newBuilder()
                    .setFileSystemPath(pythonExternalTextTransformGcsPath())
                    .setFunctionName(pythonExternalTextTransformFunctionName())
                    .build())
            .apply(
                "MapRowsToFailsafeElements",
                ParDo.of(
                        new PythonExternalTextTransformer.RowToPubSubFailsafeElementFn(
                            TRANSFORM_OUT, TRANSFORM_DEADLETTER_OUT))
                    .withOutputTags(TRANSFORM_OUT, TupleTagList.of(TRANSFORM_DEADLETTER_OUT)));
      }
      // If we don't have Python UDF, we proceed as normal checking for Javascript UDF.
      // Map the incoming messages into FailsafeElements so we can recover from failures
      // across multiple transforms.
      PCollection<FailsafeElement<PubsubMessage, String>> failsafeElements =
          input.apply("MapToRecord", ParDo.of(new PubsubMessageToFailsafeElementFn()));

      // If a Udf is supplied then use it to parse the PubSubMessages.
      if (javascriptTextTransformGcsPath() != null) {
        return failsafeElements.apply(
            "InvokeUDF",
            JavascriptTextTransformer.FailsafeJavascriptUdf.<PubsubMessage>newBuilder()
                .setFileSystemPath(javascriptTextTransformGcsPath())
                .setFunctionName(javascriptTextTransformFunctionName())
                .setReloadIntervalMinutes(javascriptTextTransformReloadIntervalMinutes())
                .setSuccessTag(TRANSFORM_OUT)
                .setFailureTag(TRANSFORM_DEADLETTER_OUT)
                .build());
      } else {
        return failsafeElements.apply(
            "ProcessPubSubMessages",
            ParDo.of(new ProcessFailsafePubSubFn())
                .withOutputTags(TRANSFORM_OUT, TupleTagList.of(TRANSFORM_DEADLETTER_OUT)));
      }
    }

    /** Builder for {@link PubSubMessageToJsonDocument}. */
    @AutoValue.Builder
    public abstract static class Builder {
      public abstract Builder setJavascriptTextTransformGcsPath(
          String javascriptTextTransformGcsPath);

      public abstract Builder setJavascriptTextTransformFunctionName(
          String javascriptTextTransformFunctionName);

      public abstract Builder setPythonExternalTextTransformGcsPath(
          String pythonExternalTextTransformGcsPath);

      public abstract Builder setPythonExternalTextTransformFunctionName(
          String pythonExternalTextTransformFunctionName);

      public abstract Builder setJavascriptTextTransformReloadIntervalMinutes(
          Integer javascriptTextTransformReloadIntervalMinutes);

      public abstract PubSubMessageToJsonDocument build();
    }
  }

  /**
   * The {@link ProcessFailsafePubSubFn} class processes a {@link FailsafeElement} containing a
   * {@link PubsubMessage} and a String of the message's payload {@link PubsubMessage#getPayload()}
   * into a {@link FailsafeElement} of the original {@link PubsubMessage} and a JSON string that has
   * been processed with {@link Gson}.
   *
   * <p>If {@link PubsubMessage#getAttributeMap()} is not empty then the message attributes will be
   * serialized along with the message payload.
   */
  static class ProcessFailsafePubSubFn
      extends DoFn<FailsafeElement<PubsubMessage, String>, FailsafeElement<PubsubMessage, String>> {

    private static final Counter successCounter =
        Metrics.counter(PubSubMessageToJsonDocument.class, "successful-json-conversion");

    private static Gson gson = new Gson();

    private static final Counter failedCounter =
        Metrics.counter(PubSubMessageToJsonDocument.class, "failed-json-conversion");

    @ProcessElement
    public void processElement(ProcessContext context) {
      PubsubMessage pubsubMessage = context.element().getOriginalPayload();

      JsonObject messageObject = new JsonObject();

      try {
        if (pubsubMessage.getPayload().length > 0) {
          messageObject =
              gson.fromJson(
                  new String(pubsubMessage.getPayload(), StandardCharsets.UTF_8), JsonObject.class);
        }

        // If message attributes are present they will be serialized along with the message payload
        if (pubsubMessage.getAttributeMap() != null) {
          pubsubMessage.getAttributeMap().forEach(messageObject::addProperty);
        }

        context.output(FailsafeElement.of(pubsubMessage, messageObject.toString()));
        successCounter.inc();

      } catch (JsonSyntaxException e) {
        context.output(
            TRANSFORM_DEADLETTER_OUT,
            FailsafeElement.of(context.element())
                .setErrorMessage(e.getMessage())
                .setStacktrace(Throwables.getStackTraceAsString(e)));
        failedCounter.inc();
      }
    }
  }

  /**
   * The {@link PubsubMessageToFailsafeElementFn} wraps an incoming {@link PubsubMessage} with the
   * {@link FailsafeElement} class so errors can be recovered from and the original message can be
   * output to a error records table.
   */
  static class PubsubMessageToFailsafeElementFn
      extends DoFn<PubsubMessage, FailsafeElement<PubsubMessage, String>> {
    @ProcessElement
    public void processElement(ProcessContext context) {
      PubsubMessage message = context.element();
      context.output(
          FailsafeElement.of(message, new String(message.getPayload(), StandardCharsets.UTF_8)));
    }
  }
}

다음 단계