Cloud Storage to Elasticsearch template

Beta

This feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

The Cloud Storage to Elasticsearch template is a batch pipeline that reads data from CSV files stored in a Cloud Storage bucket and writes the data into Elasticsearch as JSON documents.

Pipeline requirements

• The Cloud Storage bucket must exist.
• A Elasticsearch host on a Google Cloud instance or on Elasticsearch Cloud that is accessible from Dataflow must exist.
• A BigQuery table for error output must exist.

CSV schema

If the CSV files contain headers, set the containsHeaders template parameter to true.

Otherwise, create a JSON schema file that describes the data. Specify the Cloud Storage URI of the schema file in the jsonSchemaPath template parameter. The following example shows a JSON schema:

[{"name":"id", "type":"text"}, {"name":"age", "type":"integer"}]

Alternatively, you can provide a user-defined function (UDF) that parses the CSV text and outputs Elasticsearch documents.

Template parameters

Parameter	Description
inputFileSpec	The Cloud Storage file pattern to search for CSV files. Example: gs://mybucket/test-*.csv.
connectionUrl	Elasticsearch URL in the format https://hostname:[port] or specify CloudID if using Elastic Cloud.
apiKey	Base64 Encoded API key used for authentication.
index	The Elasticsearch index toward which the requests will be issued such as my-index.
deadletterTable	The BigQuery Deadletter table to send failed inserts to. Example: <your-project>:<your-dataset>.<your-table-name>.
containsHeaders	(Optional) Boolean to denote whether headers are included in the CSV. Defaultfalse.
delimiter	(Optional) The delimiter that the CSV uses. Example: ,
csvFormat	(Optional) The CSV format according to Apache Commons CSV format. Default: Default.
jsonSchemaPath	(Optional) The Path to JSON schema. Default: null.
largeNumFiles	(Optional) Set to true if number of files is in the tens of thousands. Default: false.
javascriptTextTransformGcsPath	(Optional) The Cloud Storage URI of the .js file that defines the JavaScript user-defined function (UDF) you want to use. For example, gs://my-bucket/my-udfs/my_file.js.
javascriptTextTransformFunctionName	(Optional) The name of the JavaScript user-defined function (UDF) that you want 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.
batchSize	(Optional) Batch size in number of documents. Default: 1000.
batchSizeBytes	(Optional) Batch size in number of bytes. Default: 5242880 (5mb).
maxRetryAttempts	(Optional) Max retry attempts, must be > 0. Default: no retries.
maxRetryDuration	(Optional) Max retry duration in milliseconds, must be > 0. Default: no retries.
csvFileEncoding	(Optional) CSV file encoding.
propertyAsIndex	(Optional) A property in the document being indexed whose value will specify _index metadata to be included with document in bulk request (takes precedence over an _index UDF). Default: none.
propertyAsId	(Optional) A property in the document being indexed whose value will specify _id metadata to be included with document in bulk request (takes precedence over an _id UDF). Default: none.
javaScriptIndexFnGcsPath	(Optional) The Cloud Storage path to the JavaScript UDF source for a function that will specify _index metadata to be included with document in bulk request. Default: none.
javaScriptIndexFnName	(Optional) UDF JavaScript function name for function that will specify _index metadata to be included with document in bulk request. Default: none.
javaScriptIdFnGcsPath	(Optional) The Cloud Storage path to the JavaScript UDF source for a function that will specify _id metadata to be included with document in bulk request. Default: none.
javaScriptIdFnName	(Optional) UDF JavaScript function name for function that will specify _id metadata to be included with document in bulk request. Default: none.
javaScriptTypeFnGcsPath	(Optional) The Cloud Storage path to the JavaScript UDF source for a function that will specify _type metadata to be included with document in bulk request. Default: none.
javaScriptTypeFnName	(Optional) UDF JavaScript function name for function that will specify _type metadata to be included with document in bulk request. Default: none.
javaScriptIsDeleteFnGcsPath	(Optional) The Cloud Storage path to JavaScript UDF source for function that will determine if document should be deleted rather than inserted or updated. The function should return string value "true" or "false". Default: none.
javaScriptIsDeleteFnName	(Optional) UDF JavaScript function name for function that will determine if document should be deleted rather than inserted or updated. The function should return string value "true" or "false". Default: none.
usePartialUpdate	(Optional) Whether to use partial updates (update rather than create or index, allowing partial docs) with Elasticsearch requests. Default: false.
bulkInsertMethod	(Optional) Whether to use INDEX (index, allows upserts) or CREATE (create, errors on duplicate _id) with Elasticsearch bulk requests. Default: CREATE.

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.

Text transform function

Transforms the CSV data into an Elasticsearch document.

Template parameters:

• javascriptTextTransformGcsPath: the Cloud Storage URI of the JavaScript file.
• javascriptTextTransformFunctionName: the name of the JavaScript function.

Function specification:

• Input: a single line from an input CSV file.
• Output: a stringified JSON document to insert into Elasticsearch.

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.

Caution: Mapping types were removed in Elasticsearch 8.0.

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

Console

1. Go to the Dataflow Create job from template page.
Go to Create job from template
2. In the Job name field, enter a unique job name.
3. Optional: For Regional endpoint, select a value from the drop-down menu. The default region is us-central1.

   For a list of regions where you can run a Dataflow job, see Dataflow locations.

4. From the Dataflow template drop-down menu, select the Cloud Storage to Elasticsearch template.
5. In the provided parameter fields, enter your parameter values.
6. Click Run job.

gcloud

Note: To use the Google Cloud CLI to run flex templates, you must have Google Cloud CLI version 284.0.0 or later.

In your shell or terminal, 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/GCS_to_Elasticsearch \
    --parameters \
inputFileSpec=INPUT_FILE_SPEC,\
connectionUrl=CONNECTION_URL,\
apiKey=APIKEY,\
index=INDEX,\
deadletterTable=DEADLETTER_TABLE,\

Replace the following:

• PROJECT_ID: the Google Cloud project ID where you want to run the Dataflow job
• JOB_NAME: a unique job name of your choice
• VERSION: the version of the template that you want to use

  You can use the following values:

  • latest to use the latest version of the template, which is available in the non-dated parent folder in the bucket— gs://dataflow-templates-REGION_NAME/latest/
  • the version name, like 2023-09-12-00_RC00, to use a specific version of the template, which can be found nested in the respective dated parent folder in the bucket— gs://dataflow-templates-REGION_NAME/
  Caution: The latest version of templates might update with breaking changes. Your production environments should use templates kept in the most recent dated parent folder to prevent these breaking changes from affecting your production workflows.
• REGION_NAME: the region where you want to deploy your Dataflow job—for example, us-central1
• INPUT_FILE_SPEC: your Cloud Storage file pattern.
• CONNECTION_URL: your Elasticsearch URL.
• APIKEY: your base64 encoded API key for authentication.
• INDEX: your Elasticsearch index.
• DEADLETTER_TABLE: your BigQuery table.

API

To run the template using the REST API, send an HTTP POST request. For more information on the API and its authorization scopes, see projects.templates.launch.

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

Replace the following:

• PROJECT_ID: the Google Cloud project ID where you want to run the Dataflow job
• JOB_NAME: a unique job name of your choice
• VERSION: the version of the template that you want to use

  You can use the following values:

  • latest to use the latest version of the template, which is available in the non-dated parent folder in the bucket— gs://dataflow-templates-REGION_NAME/latest/
  • the version name, like 2023-09-12-00_RC00, to use a specific version of the template, which can be found nested in the respective dated parent folder in the bucket— gs://dataflow-templates-REGION_NAME/
  Caution: The latest version of templates might update with breaking changes. Your production environments should use templates kept in the most recent dated parent folder to prevent these breaking changes from affecting your production workflows.
• LOCATION: the region where you want to deploy your Dataflow job—for example, us-central1
• INPUT_FILE_SPEC: your Cloud Storage file pattern.
• CONNECTION_URL: your Elasticsearch URL.
• APIKEY: your base64 encoded API key for authentication.
• INDEX: your Elasticsearch index.
• DEADLETTER_TABLE: your BigQuery table.

Template source code

Java

/*
 * 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 static org.apache.beam.vendor.guava.v32_1_2_jre.com.google.common.base.Preconditions.checkArgument;

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.coders.FailsafeElementCoder;
import com.google.cloud.teleport.v2.common.UncaughtExceptionLogger;
import com.google.cloud.teleport.v2.elasticsearch.options.GCSToElasticsearchOptions;
import com.google.cloud.teleport.v2.elasticsearch.transforms.WriteToElasticsearch;
import com.google.cloud.teleport.v2.transforms.CsvConverters;
import com.google.cloud.teleport.v2.transforms.ErrorConverters.WriteStringMessageErrors;
import com.google.cloud.teleport.v2.utils.SchemaUtils;
import com.google.cloud.teleport.v2.values.FailsafeElement;
import com.google.common.base.Strings;
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.NullableCoder;
import org.apache.beam.sdk.coders.StringUtf8Coder;
import org.apache.beam.sdk.options.PipelineOptionsFactory;
import org.apache.beam.sdk.transforms.MapElements;
import org.apache.beam.sdk.transforms.WithTimestamps;
import org.apache.beam.sdk.values.PCollectionTuple;
import org.apache.beam.sdk.values.TupleTag;
import org.apache.beam.sdk.values.TypeDescriptors;
import org.joda.time.Instant;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

/**
 * The {@link GCSToElasticsearch} pipeline exports data from one or more CSV files in Cloud Storage
 * to Elasticsearch.
 *
 * <p>Check out <a
 * href="https://github.com/GoogleCloudPlatform/DataflowTemplates/blob/main/v2/googlecloud-to-elasticsearch/README_GCS_to_Elasticsearch.md">README</a>
 * for instructions on how to use or modify this template.
 */
@MultiTemplate({
  @Template(
      name = "GCS_to_Elasticsearch",
      category = TemplateCategory.BATCH,
      displayName = "Cloud Storage to Elasticsearch",
      description = {
        "The Cloud Storage to Elasticsearch template is a batch pipeline that reads data from CSV files stored in a Cloud Storage bucket and writes the data into Elasticsearch as JSON documents.",
        "If the CSV files contain headers, set the <code>containsHeaders</code> template parameter to <code>true</code>.\n"
            + "Otherwise, create a JSON schema file that describes the data. Specify the Cloud Storage URI of the schema file in the jsonSchemaPath template parameter. "
            + "The following example shows a JSON schema:\n"
            + "<code>[{\"name\":\"id\", \"type\":\"text\"}, {\"name\":\"age\", \"type\":\"integer\"}]</code>\n"
            + "Alternatively, you can provide a Javascript user-defined function (UDF) that parses the CSV text and outputs Elasticsearch documents."
      },
      optionsClass = GCSToElasticsearchOptions.class,
      skipOptions = {
        "javascriptTextTransformReloadIntervalMinutes",
        "pythonExternalTextTransformGcsPath",
        "pythonExternalTextTransformFunctionName"
      },
      flexContainerName = "gcs-to-elasticsearch",
      documentation =
          "https://cloud.google.com/dataflow/docs/guides/templates/provided/cloud-storage-to-elasticsearch",
      contactInformation = "https://cloud.google.com/support",
      preview = true,
      requirements = {
        "The Cloud Storage bucket must exist.",
        "A Elasticsearch host on a Google Cloud instance or on Elasticsearch Cloud that is accessible from Dataflow must exist.",
        "A BigQuery table for error output must exist."
      }),
  @Template(
      name = "GCS_to_Elasticsearch_Xlang",
      category = TemplateCategory.BATCH,
      displayName = "Cloud Storage to Elasticsearch with Python UDFs",
      type = Template.TemplateType.XLANG,
      description = {
        "The Cloud Storage to Elasticsearch template is a batch pipeline that reads data from CSV files stored in a Cloud Storage bucket and writes the data into Elasticsearch as JSON documents.",
        "If the CSV files contain headers, set the <code>containsHeaders</code> template parameter to <code>true</code>.\n"
            + "Otherwise, create a JSON schema file that describes the data. Specify the Cloud Storage URI of the schema file in the jsonSchemaPath template parameter. "
            + "The following example shows a JSON schema:\n"
            + "<code>[{\"name\":\"id\", \"type\":\"text\"}, {\"name\":\"age\", \"type\":\"integer\"}]</code>\n"
            + "Alternatively, you can provide a Python user-defined function (UDF) that parses the CSV text and outputs Elasticsearch documents."
      },
      optionsClass = GCSToElasticsearchOptions.class,
      skipOptions = {
        "javascriptTextTransformGcsPath",
        "javascriptTextTransformFunctionName",
        "javascriptTextTransformReloadIntervalMinutes"
      },
      flexContainerName = "gcs-to-elasticsearch-xlang",
      documentation =
          "https://cloud.google.com/dataflow/docs/guides/templates/provided/cloud-storage-to-elasticsearch",
      contactInformation = "https://cloud.google.com/support",
      preview = true,
      requirements = {
        "The Cloud Storage bucket must exist.",
        "A Elasticsearch host on a Google Cloud instance or on Elasticsearch Cloud that is accessible from Dataflow must exist.",
        "A BigQuery table for error output must exist."
      })
})
public class GCSToElasticsearch {

  /** The tag for the headers of the CSV if required. */
  static final TupleTag<String> CSV_HEADERS = new TupleTag<String>() {};

  /** The tag for the lines of the CSV. */
  static final TupleTag<String> CSV_LINES = new TupleTag<String>() {};

  /** The tag for the dead-letter output of the UDF. */
  static final TupleTag<FailsafeElement<String, String>> PROCESSING_DEADLETTER_OUT =
      new TupleTag<FailsafeElement<String, String>>() {};

  /** The tag for the main output for the UDF. */
  static final TupleTag<FailsafeElement<String, String>> PROCESSING_OUT =
      new TupleTag<FailsafeElement<String, String>>() {};

  /* Logger for class. */
  private static final Logger LOG = LoggerFactory.getLogger(GCSToElasticsearch.class);

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

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

    GCSToElasticsearchOptions options =
        PipelineOptionsFactory.fromArgs(args).withValidation().as(GCSToElasticsearchOptions.class);

    run(options);
  }

  /**
   * Runs the pipeline to completion with the specified options.
   *
   * @param options The execution options.
   * @return The pipeline result.
   */
  private static PipelineResult run(GCSToElasticsearchOptions options) {
    // Create the pipeline
    Pipeline pipeline = Pipeline.create(options);

    // Register the coder for pipeline
    CoderRegistry coderRegistry = pipeline.getCoderRegistry();
    coderRegistry.registerCoderForType(
        FAILSAFE_ELEMENT_CODER.getEncodedTypeDescriptor(), FAILSAFE_ELEMENT_CODER);

    // Throw error if containsHeaders is true and a schema or Udf is also set.
    if (options.getContainsHeaders()) {
      checkArgument(
          options.getJavascriptTextTransformGcsPath() == null
              && options.getJsonSchemaPath() == null
              && options.getPythonExternalTextTransformGcsPath() == null,
          "Cannot parse file containing headers with UDF or Json schema.");
    }

    // Throw error if only one retry configuration parameter is set.
    checkArgument(
        (options.getMaxRetryAttempts() == null && options.getMaxRetryDuration() == null)
            || (options.getMaxRetryAttempts() != null && options.getMaxRetryDuration() != null),
        "To specify retry configuration both max attempts and max duration must be set.");

    // Throw error if both Javascript UDF and Python UDF are set. We can only apply one or the
    // other.
    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.");
    }

    /*
     * Steps: 1) Read records from CSV(s) via {@link CsvConverters.ReadCsv}.
     *        2) Convert lines to JSON strings via {@link CsvConverters.LineToFailsafeJson}.
     *        3a) Write JSON strings as documents to Elasticsearch via {@link ElasticsearchIO}.
     *        3b) Write elements that failed processing to {@link org.apache.beam.sdk.io.gcp.bigquery.BigQueryIO}.
     */
    PCollectionTuple readCsvLines =
        pipeline
            /*
             * Step 1: Read CSV file(s) from Cloud Storage using {@link CsvConverters.ReadCsv}.
             */
            .apply(
            "ReadCsv",
            CsvConverters.ReadCsv.newBuilder()
                .setCsvFormat(options.getCsvFormat())
                .setDelimiter(options.getDelimiter())
                .setHasHeaders(options.getContainsHeaders())
                .setInputFileSpec(options.getInputFileSpec())
                .setHeaderTag(CSV_HEADERS)
                .setLineTag(CSV_LINES)
                .setFileEncoding(options.getCsvFileEncoding())
                .build());
    /*
     * Step 2: Convert lines to Elasticsearch document.
     */
    CsvConverters.LineToFailsafeJson.Builder lineToFailsafeJsonBuilder =
        CsvConverters.LineToFailsafeJson.newBuilder()
            .setDelimiter(options.getDelimiter())
            .setJsonSchemaPath(options.getJsonSchemaPath())
            .setHeaderTag(CSV_HEADERS)
            .setLineTag(CSV_LINES)
            .setUdfOutputTag(PROCESSING_OUT)
            .setUdfDeadletterTag(PROCESSING_DEADLETTER_OUT);
    if (options.getPythonExternalTextTransformGcsPath() != null) {
      lineToFailsafeJsonBuilder
          .setPythonUdfFileSystemPath(options.getPythonExternalTextTransformGcsPath())
          .setPythonUdfFunctionName(options.getPythonExternalTextTransformFunctionName());
    } else {
      lineToFailsafeJsonBuilder
          .setJavascriptUdfFileSystemPath(options.getJavascriptTextTransformGcsPath())
          .setJavascriptUdfFunctionName(options.getJavascriptTextTransformFunctionName());
    }
    PCollectionTuple convertedCsvLines =
        readCsvLines.apply("ConvertLine", lineToFailsafeJsonBuilder.build());
    /*
     * Step 3a: Write elements that were successfully processed to Elasticsearch using {@link WriteToElasticsearch}.
     */
    convertedCsvLines
        .get(PROCESSING_OUT)
        .apply(
            "GetJsonDocuments",
            MapElements.into(TypeDescriptors.strings()).via(FailsafeElement::getPayload))
        .apply(
            "WriteToElasticsearch",
            WriteToElasticsearch.newBuilder()
                .setUserAgent("dataflow-gcs-to-elasticsearch-template/v2")
                .setOptions(options.as(GCSToElasticsearchOptions.class))
                .build());

    /*
     * Step 3b: Write elements that failed processing to deadletter table via {@link BigQueryIO}.
     */
    convertedCsvLines
        .get(PROCESSING_DEADLETTER_OUT)
        .apply(
            "AddTimestamps",
            WithTimestamps.of((FailsafeElement<String, String> failures) -> new Instant()))
        .apply(
            "WriteFailedElementsToBigQuery",
            WriteStringMessageErrors.newBuilder()
                .setErrorRecordsTable(options.getDeadletterTable())
                .setErrorRecordsTableSchema(SchemaUtils.DEADLETTER_SCHEMA)
                .build());

    return pipeline.run();
  }
}

What's next

• Learn about Dataflow templates.
• See the list of Google-provided templates.