Modelo Avro do Cloud Storage para Bigtable

O modelo Avro do Cloud Storage para Bigtable é um pipeline que lê dados de arquivos Avro em um bucket do Cloud Storage e grava os dados em uma tabela do Bigtable. É possível usar o modelo para copiar dados do Cloud Storage para o Bigtable.

Requisitos de pipeline

  • A tabela do Bigtable precisa existir e ter as mesmas famílias de colunas que foram exportadas nos arquivos Avro.
  • Os arquivos Avro de entrada precisam existir em um bucket do Cloud Storage antes de o pipeline ser executado.
  • O Bigtable espera um esquema específico dos arquivos Avro de entrada.

Parâmetros do modelo

Parâmetro Descrição
bigtableProjectId O ID do projeto Google Cloud da instância do Bigtable em que você quer gravar dados.
bigtableInstanceId O ID da instância do Bigtable que contém a tabela.
bigtableTableId O ID da tabela do Bigtable a ser importada.
inputFilePattern O padrão de caminho do Cloud Storage em que os dados estão localizados. Por exemplo, gs://mybucket/somefolder/prefix*.

Executar o modelo

  1. Acesse a página Criar job usando um modelo do Dataflow.
  2. Acesse Criar job usando um modelo
  3. No campo Nome do job, insira um nome exclusivo.
  4. Opcional: em Endpoint regional, selecione um valor no menu suspenso. A região padrão é us-central1.

    Para ver uma lista de regiões em que é possível executar um job do Dataflow, consulte Locais do Dataflow.

  5. No menu suspenso Modelo do Dataflow, selecione the Avro Files on Cloud Storage to Cloud Bigtable template.
  6. Nos campos de parâmetro fornecidos, insira os valores de parâmetro.
  7. Cliquem em Executar job.

No shell ou no terminal, execute o modelo:

gcloud dataflow jobs run JOB_NAME \
    --gcs-location gs://dataflow-templates-REGION_NAME/VERSION/GCS_Avro_to_Cloud_Bigtable \
    --region REGION_NAME \
    --parameters \
bigtableProjectId=BIGTABLE_PROJECT_ID,\
bigtableInstanceId=INSTANCE_ID,\
bigtableTableId=TABLE_ID,\
inputFilePattern=INPUT_FILE_PATTERN

Substitua:

  • JOB_NAME: um nome de job de sua escolha
  • VERSION: a versão do modelo que você quer usar

    Use estes valores:

  • REGION_NAME: a região em que você quer implantar o job do Dataflow, por exemplo, us-central1
  • BIGTABLE_PROJECT_ID: o ID do projeto Google Cloud da instância do Bigtable da qual você quer ler os dados.
  • INSTANCE_ID: o ID da instância do Bigtable que contém a tabela.
  • TABLE_ID: o ID da tabela do Cloud Bigtable a ser exportada.
  • INPUT_FILE_PATTERN: o padrão de caminho do Cloud Storage em que os dados estão localizados, por exemplo, gs://mybucket/somefolder/prefix*

Para executar o modelo usando a API REST, envie uma solicitação HTTP POST. Para mais informações sobre a API e os respectivos escopos de autorização, consulte projects.templates.launch.

POST https://dataflow.googleapis.com/v1b3/projects/PROJECT_ID/locations/LOCATION/templates:launch?gcsPath=gs://dataflow-templates-LOCATION/VERSION/GCS_Avro_to_Cloud_Bigtable
{
   "jobName": "JOB_NAME",
   "parameters": {
       "bigtableProjectId": "BIGTABLE_PROJECT_ID",
       "bigtableInstanceId": "INSTANCE_ID",
       "bigtableTableId": "TABLE_ID",
       "inputFilePattern": "INPUT_FILE_PATTERN",
   },
   "environment": { "zone": "us-central1-f" }
}

Substitua:

  • PROJECT_ID: o ID do projeto do Google Cloud em que você quer executar o job do Dataflow
  • JOB_NAME: um nome de job de sua escolha
  • VERSION: a versão do modelo que você quer usar

    Use estes valores:

  • LOCATION: a região em que você quer implantar o job do Dataflow, por exemplo, us-central1
  • BIGTABLE_PROJECT_ID: o ID do projeto Google Cloud da instância do Bigtable da qual você quer ler os dados.
  • INSTANCE_ID: o ID da instância do Bigtable que contém a tabela.
  • TABLE_ID: o ID da tabela do Cloud Bigtable a ser exportada.
  • INPUT_FILE_PATTERN: o padrão de caminho do Cloud Storage em que os dados estão localizados, por exemplo, gs://mybucket/somefolder/prefix*
Java
/*
 * Copyright (C) 2018 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.bigtable;

import com.google.bigtable.v2.Mutation;
import com.google.bigtable.v2.Mutation.SetCell;
import com.google.cloud.teleport.bigtable.AvroToBigtable.Options;
import com.google.cloud.teleport.metadata.Template;
import com.google.cloud.teleport.metadata.TemplateCategory;
import com.google.cloud.teleport.metadata.TemplateParameter;
import com.google.common.base.MoreObjects;
import com.google.common.collect.ImmutableList;
import com.google.protobuf.ByteString;
import java.nio.ByteBuffer;
import org.apache.beam.runners.dataflow.options.DataflowPipelineOptions;
import org.apache.beam.sdk.Pipeline;
import org.apache.beam.sdk.PipelineResult;
import org.apache.beam.sdk.extensions.avro.io.AvroIO;
import org.apache.beam.sdk.io.gcp.bigtable.BigtableIO;
import org.apache.beam.sdk.options.PipelineOptions;
import org.apache.beam.sdk.options.PipelineOptionsFactory;
import org.apache.beam.sdk.options.ValueProvider;
import org.apache.beam.sdk.options.ValueProvider.StaticValueProvider;
import org.apache.beam.sdk.transforms.DoFn;
import org.apache.beam.sdk.transforms.ParDo;
import org.apache.beam.sdk.values.KV;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

/**
 * Dataflow pipeline that imports data from Avro files in GCS to a Cloud Bigtable table. The Cloud
 * Bigtable table must be created before running the pipeline and must have a compatible table
 * schema. For example, if {@link BigtableCell} from the Avro files has a 'family' of "f1", the
 * Bigtable table should have a column family of "f1".
 *
 * <p>Check out <a
 * href="https://github.com/GoogleCloudPlatform/DataflowTemplates/blob/main/v1/README_GCS_Avro_to_Cloud_Bigtable.md">README</a>
 * for instructions on how to use or modify this template.
 */
@Template(
    name = "GCS_Avro_to_Cloud_Bigtable",
    category = TemplateCategory.BATCH,
    displayName = "Avro Files on Cloud Storage to Cloud Bigtable",
    description =
        "The Cloud Storage Avro to Bigtable template is a pipeline that reads data from Avro files in a Cloud Storage bucket and writes the data to a Bigtable table. "
            + "You can use the template to copy data from Cloud Storage to Bigtable.",
    optionsClass = Options.class,
    documentation =
        "https://cloud.google.com/dataflow/docs/guides/templates/provided/avro-to-bigtable",
    contactInformation = "https://cloud.google.com/support",
    requirements = {
      "The Bigtable table must exist and have the same column families as exported in the Avro files.",
      "The input Avro files must exist in a Cloud Storage bucket before running the pipeline.",
      "Bigtable expects a specific <a href=\"https://github.com/GoogleCloudPlatform/DataflowTemplates/blob/main/v1/src/main/resources/schema/avro/bigtable.avsc\">schema</a> from the input Avro files.",
    })
public final class AvroToBigtable {
  private static final Logger LOG = LoggerFactory.getLogger(AvroToBigtable.class);

  /** Maximum number of mutations allowed per row by Cloud bigtable. */
  private static final int MAX_MUTATIONS_PER_ROW = 100000;

  private static final Boolean DEFAULT_SPLIT_LARGE_ROWS = false;

  /** Options for the import pipeline. */
  public interface Options extends PipelineOptions {
    @TemplateParameter.ProjectId(
        order = 1,
        groupName = "Target",
        description = "Project ID",
        helpText =
            "The ID of the Google Cloud project that contains the Bigtable instance that you want to write data to.")
    ValueProvider<String> getBigtableProjectId();

    @SuppressWarnings("unused")
    void setBigtableProjectId(ValueProvider<String> projectId);

    @TemplateParameter.Text(
        order = 2,
        groupName = "Target",
        regexes = {"[a-z][a-z0-9\\-]+[a-z0-9]"},
        description = "Instance ID",
        helpText = "The ID of the Bigtable instance that contains the table.")
    ValueProvider<String> getBigtableInstanceId();

    @SuppressWarnings("unused")
    void setBigtableInstanceId(ValueProvider<String> instanceId);

    @TemplateParameter.Text(
        order = 4,
        groupName = "Target",
        regexes = {"[_a-zA-Z0-9][-_.a-zA-Z0-9]*"},
        description = "Table ID",
        helpText = "The ID of the Bigtable table to import.")
    ValueProvider<String> getBigtableTableId();

    @SuppressWarnings("unused")
    void setBigtableTableId(ValueProvider<String> tableId);

    @TemplateParameter.GcsReadFile(
        order = 5,
        groupName = "Source",
        description = "Input Cloud Storage File(s)",
        helpText = "The Cloud Storage path pattern where data is located.",
        example = "gs://<BUCKET_NAME>/FOLDER/PREFIX*")
    ValueProvider<String> getInputFilePattern();

    @SuppressWarnings("unused")
    void setInputFilePattern(ValueProvider<String> inputFilePattern);

    @TemplateParameter.Boolean(
        order = 6,
        optional = true,
        description = "If true, large rows will be split into multiple MutateRows requests",
        helpText =
            "The flag for enabling splitting of large rows into multiple MutateRows requests. Note that when a large row is split between multiple API calls, the updates to the row are not atomic.")
    ValueProvider<Boolean> getSplitLargeRows();

    void setSplitLargeRows(ValueProvider<Boolean> splitLargeRows);
  }

  /**
   * Runs a pipeline to import Avro files in GCS to a Cloud Bigtable table.
   *
   * @param args arguments to the pipeline
   */
  public static void main(String[] args) {
    Options options = PipelineOptionsFactory.fromArgs(args).withValidation().as(Options.class);
    PipelineResult result = run(options);

    // Wait for pipeline to finish only if it is not constructing a template.
    if (options.as(DataflowPipelineOptions.class).getTemplateLocation() == null) {
      result.waitUntilFinish();
    }
  }

  public static PipelineResult run(Options options) {
    Pipeline pipeline = Pipeline.create(PipelineUtils.tweakPipelineOptions(options));

    BigtableIO.Write write =
        BigtableIO.write()
            .withProjectId(options.getBigtableProjectId())
            .withInstanceId(options.getBigtableInstanceId())
            .withTableId(options.getBigtableTableId());

    pipeline
        .apply("Read from Avro", AvroIO.read(BigtableRow.class).from(options.getInputFilePattern()))
        .apply(
            "Transform to Bigtable",
            ParDo.of(
                AvroToBigtableFn.createWithSplitLargeRows(
                    options.getSplitLargeRows(), MAX_MUTATIONS_PER_ROW)))
        .apply("Write to Bigtable", write);

    return pipeline.run();
  }

  /**
   * Translates {@link BigtableRow} to {@link Mutation}s along with a row key. The mutations are
   * {@link SetCell}s that set the value for specified cells with family name, column qualifier and
   * timestamp.
   */
  static class AvroToBigtableFn extends DoFn<BigtableRow, KV<ByteString, Iterable<Mutation>>> {
    private final ValueProvider<Boolean> splitLargeRowsFlag;
    private Boolean splitLargeRows;
    private final int maxMutationsPerRow;

    public static AvroToBigtableFn create() {
      return new AvroToBigtableFn(StaticValueProvider.of(false), MAX_MUTATIONS_PER_ROW);
    }

    public static AvroToBigtableFn createWithSplitLargeRows(
        ValueProvider<Boolean> splitLargeRowsFlag, int maxMutationsPerRequest) {
      return new AvroToBigtableFn(splitLargeRowsFlag, maxMutationsPerRequest);
    }

    private AvroToBigtableFn(
        ValueProvider<Boolean> splitLargeRowsFlag, int maxMutationsPerRequest) {
      this.splitLargeRowsFlag = splitLargeRowsFlag;
      this.maxMutationsPerRow = maxMutationsPerRequest;
    }

    @Setup
    public void setup() {
      if (splitLargeRowsFlag != null) {
        splitLargeRows = splitLargeRowsFlag.get();
      }
      splitLargeRows = MoreObjects.firstNonNull(splitLargeRows, DEFAULT_SPLIT_LARGE_ROWS);
      LOG.info("splitLargeRows set to: " + splitLargeRows);
    }

    @ProcessElement
    public void processElement(
        @Element BigtableRow row, OutputReceiver<KV<ByteString, Iterable<Mutation>>> out) {
      ByteString key = toByteString(row.getKey());
      // BulkMutation doesn't split rows. Currently, if a single row contains more than 100,000
      // mutations, the service will fail the request.
      ImmutableList.Builder<Mutation> mutations = ImmutableList.builder();
      int cellsProcessed = 0;
      for (BigtableCell cell : row.getCells()) {
        SetCell setCell =
            SetCell.newBuilder()
                .setFamilyName(cell.getFamily().toString())
                .setColumnQualifier(toByteString(cell.getQualifier()))
                .setTimestampMicros(cell.getTimestamp())
                .setValue(toByteString(cell.getValue()))
                .build();

        mutations.add(Mutation.newBuilder().setSetCell(setCell).build());
        cellsProcessed++;

        if (this.splitLargeRows && cellsProcessed % maxMutationsPerRow == 0) {
          // Send a MutateRow request when we have accumulated max mutations per row.
          out.output(KV.of(key, mutations.build()));
          mutations = ImmutableList.builder();
        }
      }

      // Flush any remaining mutations.
      ImmutableList remainingMutations = mutations.build();
      if (!remainingMutations.isEmpty()) {
        out.output(KV.of(key, remainingMutations));
      }
    }
  }

  /** Copies the content in {@code byteBuffer} into a {@link ByteString}. */
  protected static ByteString toByteString(ByteBuffer byteBuffer) {
    return ByteString.copyFrom(byteBuffer.array());
  }
}

A seguir