BigQuery to Elasticsearch template

The BigQuery to Elasticsearch template is a batch pipeline that ingests data from a BigQuery table into Elasticsearch as documents. The template can either read the entire table or read specific records using a supplied query.

Pipeline requirements

• The source BigQuery table must exist.
• A Elasticsearch host on a Google Cloud instance or on Elastic Cloud with Elasticsearch version 7.0 or later. Must be accessible from the Dataflow worker machines.

Template parameters

Required parameters

• connectionUrl : The Elasticsearch URL in the format https://hostname:[port]. If using Elastic Cloud, specify the CloudID. (Example: https://elasticsearch-host:9200).
• apiKey : The Base64-encoded API key to use for authentication.
• index : The Elasticsearch index that the requests are issued to, such as my-index. (Example: my-index).

Optional parameters

• inputTableSpec : The BigQuery table to read from. Format: projectId:datasetId.tablename. If you specify inputTableSpec, the template reads the data directly from BigQuery storage by using the BigQuery Storage Read API (https://cloud.google.com/bigquery/docs/reference/storage). For information about limitations in the Storage Read API, see https://cloud.google.com/bigquery/docs/reference/storage#limitations. You must specify either inputTableSpec or query. If you set both parameters, the template uses the query parameter. (Example: bigquery-project:dataset.input_table).
• outputDeadletterTable : The BigQuery table for messages that failed to reach the output table, in the format <PROJECT_ID>:<DATASET_NAME>.<DEADLETTER_TABLE>. If a table doesn't exist, is is created during pipeline execution. If not specified, <outputTableSpec>_error_records is used. (Example: your-project-id:your-dataset.your-table-name).
• query : The SQL query to use to read data from BigQuery. If the BigQuery dataset is in a different project than the Dataflow job, specify the full dataset name in the SQL query, for example: <PROJECT_ID>.<DATASET_NAME>.<TABLE_NAME>. By default, the query parameter uses GoogleSQL (https://cloud.google.com/bigquery/docs/introduction-sql), unless useLegacySql is true. You must specify either inputTableSpec or query. If you set both parameters, the template uses the query parameter. (Example: select * from sampledb.sample_table).
• useLegacySql : Set to true to use legacy SQL. This parameter only applies when using the query parameter. Defaults to: false.
• queryLocation : Needed when reading from an authorized view without underlying table's permission. (Example: US).
• elasticsearchUsername : The Elasticsearch username to authenticate with. If specified, the value of 'apiKey' is ignored.
• elasticsearchPassword : The Elasticsearch password to authenticate with. If specified, the value of 'apiKey' is ignored.
• batchSize : The batch size in number of documents. Defaults to: 1000.
• batchSizeBytes : The batch size in number of bytes. Defaults to: 5242880 (5mb).
• maxRetryAttempts : The maximum number of retry attempts. Must be greater than zero. Defaults to: no retries.
• maxRetryDuration : The maximum retry duration in milliseconds. Must be greater than zero. Defaults to: no retries.
• propertyAsIndex : The property in the document being indexed whose value specifies _index metadata to include with the document in bulk requests. Takes precedence over an _index UDF. Defaults to: none.
• javaScriptIndexFnGcsPath : The Cloud Storage path to the JavaScript UDF source for a function that specifies _index metadata to include with the document in bulk requests. Defaults to: none.
• javaScriptIndexFnName : The name of the UDF JavaScript function that specifies _index metadata to include with the document in bulk requests. Defaults to: none.
• propertyAsId : A property in the document being indexed whose value specifies _id metadata to include with the document in bulk requests. Takes precedence over an _id UDF. Defaults to: none.
• javaScriptIdFnGcsPath : The Cloud Storage path to the JavaScript UDF source for the function that specifies _id metadata to include with the document in bulk requests. Defaults to: none.
• javaScriptIdFnName : The name of the UDF JavaScript function that specifies the _id metadata to include with the document in bulk requests. Defaults to: none.
• javaScriptTypeFnGcsPath : The Cloud Storage path to the JavaScript UDF source for a function that specifies _type metadata to include with documents in bulk requests. Default: none.
• javaScriptTypeFnName : The name of the UDF JavaScript function that specifies the _type metadata to include with the document in bulk requests. Defaults to: none.
• javaScriptIsDeleteFnGcsPath : The Cloud Storage path to the JavaScript UDF source for the function that determines whether to delete the document instead of inserting or updating it. The function returns a string value of true or false. Defaults to: none.
• javaScriptIsDeleteFnName : The name of the UDF JavaScript function that determines whether to delete the document instead of inserting or updating it. The function returns a string value of true or false. Defaults to: none.
• usePartialUpdate : Whether to use partial updates (update rather than create or index, allowing partial documents) with Elasticsearch requests. Defaults to: false.
• bulkInsertMethod : Whether to use INDEX (index, allows upserts) or CREATE (create, errors on duplicate _id) with Elasticsearch bulk requests. Defaults to: CREATE.
• trustSelfSignedCerts : Whether to trust self-signed certificate or not. An Elasticsearch instance installed might have a self-signed certificate, Enable this to True to by-pass the validation on SSL certificate. (default is False).
• disableCertificateValidation : If 'true', trust the self-signed SSL certificate. An Elasticsearch instance might have a self-signed certificate. To bypass validation for the certificate, set this parameter to 'true'. Default: false.
• apiKeyKMSEncryptionKey : The Cloud KMS key to decrypt the API key. This parameter must be provided if the apiKeySource is set to KMS. If this parameter is provided, apiKey string should be passed in encrypted. Encrypt parameters using the KMS API encrypt endpoint. The Key should be in the format projects/{gcp_project}/locations/{key_region}/keyRings/{key_ring}/cryptoKeys/{kms_key_name}. See: https://cloud.google.com/kms/docs/reference/rest/v1/projects.locations.keyRings.cryptoKeys/encrypt (Example: projects/your-project-id/locations/global/keyRings/your-keyring/cryptoKeys/your-key-name).
• apiKeySecretId : Secret Manager secret ID for the apiKey. This parameter should be provided if the apiKeySource is set to SECRET_MANAGER. Should be in the format projects/{project}/secrets/{secret}/versions/{secret_version}. (Example: projects/your-project-id/secrets/your-secret/versions/your-secret-version).
• apiKeySource : Source of the API key. One of PLAINTEXT, KMS or SECRET_MANAGER. This parameter must be provided if secret manager or KMS is used. If apiKeySource is set to KMS, apiKeyKMSEncryptionKey and encrypted apiKey must be provided. If apiKeySource is set to SECRET_MANAGER, apiKeySecretId must be provided. If apiKeySource is set to PLAINTEXT, apiKey must be provided. Defaults to: PLAINTEXT.
• javascriptTextTransformGcsPath : The Cloud Storage URI of the .js file that defines the JavaScript user-defined function (UDF) to use. (Example: gs://my-bucket/my-udfs/my_file.js).
• javascriptTextTransformFunctionName : The name of the JavaScript user-defined function (UDF) to use. For example, if your JavaScript function code is myTransform(inJson) { /*...do stuff...*/ }, then the function name is myTransform. For sample JavaScript UDFs, see UDF Examples (https://github.com/GoogleCloudPlatform/DataflowTemplates#udf-examples).

User-defined functions

This template supports user-defined functions (UDFs) at several points in the pipeline, described below. For more information, see Create user-defined functions for Dataflow templates.

Index function

Returns the index to which the document belongs.

Template parameters:

• javaScriptIndexFnGcsPath: the Cloud Storage URI of the JavaScript file.
• javaScriptIndexFnName: the name of the JavaScript function.

Function specification:

• Input: the Elasticsearch document, serialized as a JSON string.
• Output: the value of the document's _index metadata field.

Document ID function

Returns the document ID.

Template parameters:

• javaScriptIdFnGcsPath: the Cloud Storage URI of the JavaScript file.
• javaScriptIdFnName: the name of the JavaScript function.

Function specification:

• Input: the Elasticsearch document, serialized as a JSON string.
• Output: the value of the document's _id metadata field.

Document deletion function

Specifies whether to delete a document. To use this function, set the bulk insert mode to INDEX and provide a document ID function.

Template parameters:

• javaScriptIsDeleteFnGcsPath: the Cloud Storage URI of the JavaScript file.
• javaScriptIsDeleteFnName: the name of the JavaScript function.

Function specification:

• Input: the Elasticsearch document, serialized as a JSON string.
• Output: return the string "true" to delete the document, or "false" to upsert the document.

Mapping type function

Returns the document's mapping type.

Template parameters:

• javaScriptTypeFnGcsPath: the Cloud Storage URI of the JavaScript file.
• javaScriptTypeFnName: the name of the JavaScript function.

Function specification:

• Input: the Elasticsearch document, serialized as a JSON string.
• Output: the value of the document's _type metadata field.

Run the template

gcloud dataflow flex-template run JOB_NAME \
    --project=PROJECT_ID \
    --region=REGION_NAME \
    --template-file-gcs-location=gs://dataflow-templates-REGION_NAME/VERSION/flex/BigQuery_to_Elasticsearch \
    --parameters \
inputTableSpec=INPUT_TABLE_SPEC,\
connectionUrl=CONNECTION_URL,\
apiKey=APIKEY,\
index=INDEX

POST https://dataflow.googleapis.com/v1b3/projects/PROJECT_ID/locations/LOCATION/flexTemplates:launch
{
   "launch_parameter": {
      "jobName": "JOB_NAME",
      "parameters": {
          "inputTableSpec": "INPUT_TABLE_SPEC",
          "connectionUrl": "CONNECTION_URL",
          "apiKey": "APIKEY",
          "index": "INDEX"
      },
      "containerSpecGcsPath": "gs://dataflow-templates-LOCATION/VERSION/flex/BigQuery_to_Elasticsearch",
   }
}

/*
 * Copyright (C) 2021 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.elasticsearch.templates;

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.v2.common.UncaughtExceptionLogger;
import com.google.cloud.teleport.v2.elasticsearch.options.BigQueryToElasticsearchOptions;
import com.google.cloud.teleport.v2.elasticsearch.transforms.WriteToElasticsearch;
import com.google.cloud.teleport.v2.transforms.BigQueryConverters.ReadBigQueryTableRows;
import com.google.cloud.teleport.v2.transforms.BigQueryConverters.TableRowToJsonFn;
import com.google.cloud.teleport.v2.transforms.JavascriptTextTransformer.TransformTextViaJavascript;
import com.google.cloud.teleport.v2.transforms.PythonExternalTextTransformer;
import com.google.common.base.Strings;
import org.apache.beam.sdk.Pipeline;
import org.apache.beam.sdk.PipelineResult;
import org.apache.beam.sdk.options.PipelineOptionsFactory;
import org.apache.beam.sdk.transforms.ParDo;
import org.apache.beam.sdk.values.PCollection;

/**
 * The {@link BigQueryToElasticsearch} pipeline exports data from a BigQuery table to Elasticsearch.
 *
 * <p>Check out <a
 * href="https://github.com/GoogleCloudPlatform/DataflowTemplates/blob/main/v2/googlecloud-to-elasticsearch/README_BigQuery_to_Elasticsearch.md">README</a>
 * for instructions on how to use or modify this template.
 */
@MultiTemplate({
  @Template(
      name = "BigQuery_to_Elasticsearch",
      category = TemplateCategory.BATCH,
      displayName = "BigQuery to Elasticsearch",
      description =
          "The BigQuery to Elasticsearch template is a batch pipeline that ingests data from a BigQuery table into Elasticsearch as documents. "
              + "The template can either read the entire table or read specific records using a supplied query.",
      optionsClass = BigQueryToElasticsearchOptions.class,
      skipOptions = {
        "javascriptTextTransformReloadIntervalMinutes",
        "pythonExternalTextTransformGcsPath",
        "pythonExternalTextTransformFunctionName"
      },
      flexContainerName = "bigquery-to-elasticsearch",
      documentation =
          "https://cloud.google.com/dataflow/docs/guides/templates/provided/bigquery-to-elasticsearch",
      contactInformation = "https://cloud.google.com/support",
      preview = true,
      requirements = {
        "The source BigQuery table must exist.",
        "A Elasticsearch host on a Google Cloud instance or on Elastic Cloud with Elasticsearch version 7.0 or above and should be accessible from the Dataflow worker machines.",
      }),
  @Template(
      name = "BigQuery_to_Elasticsearch_Xlang",
      category = TemplateCategory.BATCH,
      displayName = "BigQuery to Elasticsearch with Python UDFs",
      type = Template.TemplateType.XLANG,
      description =
          "The BigQuery to Elasticsearch template is a batch pipeline that ingests data from a BigQuery table into Elasticsearch as documents. "
              + "The template can either read the entire table or read specific records using a supplied query.",
      optionsClass = BigQueryToElasticsearchOptions.class,
      skipOptions = {
        "javascriptTextTransformReloadIntervalMinutes",
        "javascriptTextTransformGcsPath",
        "javascriptTextTransformFunctionName"
      },
      flexContainerName = "bigquery-to-elasticsearch-xlang",
      documentation =
          "https://cloud.google.com/dataflow/docs/guides/templates/provided/bigquery-to-elasticsearch",
      contactInformation = "https://cloud.google.com/support",
      preview = true,
      requirements = {
        "The source BigQuery table must exist.",
        "A Elasticsearch host on a Google Cloud instance or on Elastic Cloud with Elasticsearch version 7.0 or above and should be accessible from the Dataflow worker machines.",
      })
})
public class BigQueryToElasticsearch {
  /**
   * Main entry point for pipeline execution.
   *
   * @param args Command line arguments to the pipeline.
   */
  public static void main(String[] args) {
    UncaughtExceptionLogger.register();

    BigQueryToElasticsearchOptions options =
        PipelineOptionsFactory.fromArgs(args)
            .withValidation()
            .as(BigQueryToElasticsearchOptions.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.
   */
  private static PipelineResult run(BigQueryToElasticsearchOptions options) {

    // Create the pipeline.
    Pipeline pipeline = Pipeline.create(options);
    /*
     * Steps: 1) Read records from BigQuery via BigQueryIO.
     *        2) Create json string from Table Row.
     *        3) Write records to Elasticsearch.
     */

    boolean useJavascriptUdf = !Strings.isNullOrEmpty(options.getJavascriptTextTransformGcsPath());
    boolean usePythonUdf = !Strings.isNullOrEmpty(options.getPythonExternalTextTransformGcsPath());
    if (useJavascriptUdf && usePythonUdf) {
      throw new IllegalArgumentException(
          "Either javascript or Python gcs path must be provided, but not both.");
    }

    /*
     * Step #1: Read from BigQuery. If a query is provided then it is used to get the TableRows.
     */
    PCollection<String> readJsonDocuments =
        pipeline
            .apply(
                "ReadFromBigQuery",
                ReadBigQueryTableRows.newBuilder()
                    .setOptions(options.as(BigQueryToElasticsearchOptions.class))
                    .build())

            /*
             * Step #2: Convert table rows to JSON documents.
             */
            .apply("TableRowsToJsonDocument", ParDo.of(new TableRowToJsonFn()));

    /*
     * Step #3: Apply UDF functions (if specified)
     */
    PCollection<String> udfOut;
    if (usePythonUdf) {
      udfOut =
          readJsonDocuments
              .apply(
                  "MapToRecord",
                  PythonExternalTextTransformer.FailsafeRowPythonExternalUdf
                      .stringMappingFunction())
              .setRowSchema(PythonExternalTextTransformer.FailsafeRowPythonExternalUdf.ROW_SCHEMA)
              .apply(
                  "InvokeUDF",
                  PythonExternalTextTransformer.FailsafePythonExternalUdf.newBuilder()
                      .setFileSystemPath(options.getPythonExternalTextTransformGcsPath())
                      .setFunctionName(options.getPythonExternalTextTransformFunctionName())
                      .build())
              .apply(
                  "MapToStringElements",
                  ParDo.of(new PythonExternalTextTransformer.RowToStringElementFn()));
    } else {
      udfOut =
          readJsonDocuments.apply(
              TransformTextViaJavascript.newBuilder()
                  .setFileSystemPath(options.getJavascriptTextTransformGcsPath())
                  .setFunctionName(options.getJavascriptTextTransformFunctionName())
                  .build());
    }

    /*
     * Step #4: Write converted records to Elasticsearch
     */
    udfOut.apply(
        "WriteToElasticsearch",
        WriteToElasticsearch.newBuilder()
            .setUserAgent("dataflow-bigquery-to-elasticsearch-template/v2")
            .setOptions(options.as(BigQueryToElasticsearchOptions.class))
            .build());

    return pipeline.run();
  }
}