Per utilizzare il connettore TextIO o AvroIO con Cloud Storage, includi
la seguente dipendenza. Questa libreria fornisce un gestore dello schema per i nomi dei file "gs://".
Abilitare gRPC sul connettore I/O Apache Beam su Dataflow
Puoi connetterti a Cloud Storage utilizzando gRPC tramite il
connettore I/O Apache Beam su Dataflow. gRPC è un
framework open source per chiamata di procedura remota (RPC) ad alte prestazioni sviluppato da
Google che puoi utilizzare per interagire con
Cloud Storage.
Per velocizzare le richieste di lettura del job Dataflow a Cloud Storage, puoi attivare il connettore I/O Apache Beam su Dataflow per utilizzare gRPC.
Puoi configurare il connettore I/O Apache Beam su Dataflow per generare metriche correlate a gRPC in Cloud Monitoring. Le metriche correlate a gRPC possono aiutarti a:
Monitora e ottimizza le prestazioni delle richieste gRPC a Cloud Storage.
Risolvi i problemi ed esegui il debug.
Ottieni informazioni sull'utilizzo e sul comportamento della tua applicazione.
Per informazioni su come configurare il connettore Apache Beam I/O su Dataflow per generare metriche correlate a gRPC, consulta Utilizzare le metriche lato client.
Se la raccolta di metriche non è necessaria per il tuo caso d'uso, puoi disattivarla.
Per istruzioni, vedi Disattivare le metriche lato client.
Parallelismo
I connettori TextIO e AvroIO supportano due livelli di parallelismo:
I singoli file vengono codificati separatamente, in modo che possano essere letti da più worker.
Se i file non sono compressi, il connettore può leggere separatamente i sottointervalli di ogni file, il che porta a un livello di parallelismo molto elevato. Questa suddivisione è
possibile solo se ogni riga del file è un record significativo. Ad esempio,
non è disponibile per impostazione predefinita per i file JSON.
Prestazioni
La tabella seguente mostra le metriche di rendimento per la lettura da Cloud Storage. I carichi di lavoro sono stati eseguiti su un worker e2-standard2,
utilizzando l'SDK Apache Beam 2.49.0 per Java. Non hanno utilizzato Runner v2.
Queste metriche si basano su semplici pipeline batch. Sono pensati per confrontare le prestazioni
tra i connettori I/O e non sono necessariamente rappresentativi delle pipeline del mondo reale.
Il rendimento della pipeline Dataflow è complesso e dipende dal tipo di VM, dai dati
in fase di elaborazione, dal rendimento delle origini e dei sink esterni e dal codice utente. Le metriche si basano
sull'esecuzione dell'SDK Java e non sono rappresentative delle caratteristiche di prestazioni di altri
SDK di linguaggio. Per ulteriori informazioni, vedi Prestazioni
di Beam IO.
Best practice
Evita di utilizzare watchForNewFiles con
Cloud Storage. Questo approccio non è scalabile per pipeline di produzione di grandi dimensioni, perché il connettore deve conservare un elenco dei file visualizzati in memoria. L'elenco non può essere svuotato dalla memoria, il che riduce la memoria di lavoro dei worker nel tempo. Ti consigliamo di utilizzare le
notifiche Pub/Sub per Cloud Storage. Per saperne di più, consulta
Pattern di elaborazione dei file.
Se sia il nome file che i contenuti del file sono dati utili, utilizza la classe
FileIO per leggere i nomi file. Ad esempio, un nome file potrebbe
contenere metadati utili per l'elaborazione dei dati nel file. Per ulteriori
informazioni, vedi
Accesso ai nomi dei file.
La documentazione di FileIO mostra anche un esempio di questo pattern.
Esempio
L'esempio seguente mostra come leggere da Cloud Storage.
importorg.apache.beam.sdk.Pipeline;importorg.apache.beam.sdk.PipelineResult;importorg.apache.beam.sdk.io.TextIO;importorg.apache.beam.sdk.options.Description;importorg.apache.beam.sdk.options.PipelineOptions;importorg.apache.beam.sdk.options.PipelineOptionsFactory;importorg.apache.beam.sdk.transforms.MapElements;importorg.apache.beam.sdk.values.TypeDescriptors;publicclassReadFromStorage{publicstaticPipelinecreatePipeline(Optionsoptions){varpipeline=Pipeline.create(options);pipeline// Read from a text file..apply(TextIO.read().from("gs://"+options.getBucket()+"/*.txt")).apply(MapElements.into(TypeDescriptors.strings()).via((x->{System.out.println(x);returnx;})));returnpipeline;}}
[[["Facile da capire","easyToUnderstand","thumb-up"],["Il problema è stato risolto","solvedMyProblem","thumb-up"],["Altra","otherUp","thumb-up"]],[["Difficile da capire","hardToUnderstand","thumb-down"],["Informazioni o codice di esempio errati","incorrectInformationOrSampleCode","thumb-down"],["Mancano le informazioni o gli esempi di cui ho bisogno","missingTheInformationSamplesINeed","thumb-down"],["Problema di traduzione","translationIssue","thumb-down"],["Altra","otherDown","thumb-down"]],["Ultimo aggiornamento 2025-09-04 UTC."],[[["\u003cp\u003eTo read data from Cloud Storage to Dataflow, use the Apache Beam \u003ccode\u003eTextIO\u003c/code\u003e or \u003ccode\u003eAvroIO\u003c/code\u003e I/O connector and include the Google Cloud library dependency, which provides a schema handler for \u003ccode\u003e"gs://"\u003c/code\u003e filenames.\u003c/p\u003e\n"],["\u003cp\u003eEnabling gRPC through the Apache Beam I/O connector on Dataflow can accelerate Dataflow job read requests to Cloud Storage, using the pipeline option \u003ccode\u003e--additional-experiments=use_grpc_for_gcs\u003c/code\u003e or \u003ccode\u003e--experiments=use_grpc_for_gcs\u003c/code\u003e, and requires Apache Beam SDK version 2.55.0 or later.\u003c/p\u003e\n"],["\u003cp\u003eThe \u003ccode\u003eTextIO\u003c/code\u003e and \u003ccode\u003eAvroIO\u003c/code\u003e connectors offer parallelism by allowing multiple workers to read individual files separately, and for uncompressed files, sub-ranges of each file can be read separately, enhancing parallelism.\u003c/p\u003e\n"],["\u003cp\u003eAvoid using \u003ccode\u003ewatchForNewFiles\u003c/code\u003e with Cloud Storage in large production pipelines, and use Pub/Sub notifications instead to prevent memory issues, and use \u003ccode\u003eFileIO\u003c/code\u003e class when both the filename and file contents are valuable data.\u003c/p\u003e\n"],["\u003cp\u003ePerformance metrics for reading from Cloud Storage using the Apache Beam SDK 2.49.0 for Java on an \u003ccode\u003ee2-standard2\u003c/code\u003e worker showed throughputs of 320 MBps and 320,000 elements per second for 100M records with 1 kB and 1 column.\u003c/p\u003e\n"]]],[],null,["# Read from Cloud Storage to Dataflow\n\nTo read data from Cloud Storage to Dataflow, use the\nApache Beam `TextIO` or `AvroIO`\n[I/O connector](https://beam.apache.org/documentation/io/connectors/).\n| **Note:** Depending on your scenario, consider using one of the [Google-provided Dataflow templates](/dataflow/docs/guides/templates/provided-templates). Several of these templates read from Cloud Storage.\n\nInclude the Google Cloud library dependency\n-------------------------------------------\n\nTo use the `TextIO` or `AvroIO` connector with Cloud Storage, include\nthe following dependency. This library provides a schema handler for `\"gs://\"`\nfilenames. \n\n### Java\n\n \u003cdependency\u003e\n \u003cgroupId\u003eorg.apache.beam\u003c/groupId\u003e\n \u003cartifactId\u003ebeam-sdks-java-io-google-cloud-platform\u003c/artifactId\u003e\n \u003cversion\u003e${beam.version}\u003c/version\u003e\n \u003c/dependency\u003e\n\n### Python\n\n apache-beam[gcp]==\u003cvar translate=\"no\"\u003e\u003cspan class=\"devsite-syntax-n\"\u003eVERSION\u003c/span\u003e\u003c/var\u003e\n\n### Go\n\n import _ \"github.com/apache/beam/sdks/v2/go/pkg/beam/io/filesystem/gcs\"\n\nFor more information, see\n[Install the Apache Beam SDK](/dataflow/docs/guides/installing-beam-sdk).\n\nEnable gRPC on Apache Beam I/O connector on Dataflow\n----------------------------------------------------\n\nYou can [connect to Cloud Storage using gRPC](/storage/docs/enable-grpc-api) through the\nApache Beam I/O connector on Dataflow. [gRPC](https://grpc.io/) is a\nhigh performance open-source remote procedure call (RPC) framework developed by\nGoogle that you can use to interact with\nCloud Storage.\n\nTo speed up your Dataflow job's read requests to Cloud Storage, you\ncan enable the Apache Beam I/O connector on Dataflow to use gRPC.\n\n### Command line\n\n1. Ensure that you use the [Apache Beam SDK](https://beam.apache.org/documentation/programming-guide/#configuring-pipeline-options) version [2.55.0](https://beam.apache.org/get-started/beam-overview/) or later.\n2. To run a Dataflow job, use `--additional-experiments=use_grpc_for_gcs` pipeline option. For information about the different pipeline options, see [Optional flags](/sdk/gcloud/reference/dataflow/jobs/run#--additional-experiments).\n\n### Apache Beam SDK\n\n1. Ensure that you use the [Apache Beam SDK](https://beam.apache.org/documentation/programming-guide/#configuring-pipeline-options) version [2.55.0](https://beam.apache.org/get-started/beam-overview/) or later.\n2. To run a Dataflow job, use `--experiments=use_grpc_for_gcs` pipeline option. For information about the different pipeline options, see [Basic\n options](/dataflow/docs/reference/pipeline-options#basic_options).\n\nYou can configure Apache Beam I/O connector on Dataflow to generate gRPC\nrelated metrics in Cloud Monitoring. The gRPC related metrics can help you to do the following:\n\n- Monitor and optimize the performance of gRPC requests to Cloud Storage.\n- Troubleshoot and debug issues.\n- Gain insights into your application's usage and behavior.\n\n\u003cbr /\u003e\n\nFor information about how to configure Apache Beam I/O connector on Dataflow to generate gRPC related metrics, see [Use client-side metrics](/storage/docs/client-side-metrics). If gathering metrics isn't necessary for your use case, you can choose to opt-out of metrics collection. For instructions, see [Opt-out of client-side\nmetrics](/storage/docs/client-side-metrics#opt_out_of_client-side_metrics).\n\n\u003cbr /\u003e\n\nParallelism\n-----------\n\nThe `TextIO` and `AvroIO` connectors support two levels of parallelism:\n\n- Individual files are keyed separately, so that multiple workers can read them.\n- If the files are uncompressed, the connector can read sub-ranges of each file separately, leading to a very high level of parallelism. This splitting is only possible if each line in the file is a meaningful record. For example, it's not available by default for JSON files.\n\nPerformance\n-----------\n\nThe following table shows performance metrics for reading from\nCloud Storage. The workloads were run on one `e2-standard2` worker,\nusing the Apache Beam SDK 2.49.0 for Java. They did not use Runner v2.\n\n\nThese metrics are based on simple batch pipelines. They are intended to compare performance\nbetween I/O connectors, and are not necessarily representative of real-world pipelines.\nDataflow pipeline performance is complex, and is a function of VM type, the data\nbeing processed, the performance of external sources and sinks, and user code. Metrics are based\non running the Java SDK, and aren't representative of the performance characteristics of other\nlanguage SDKs. For more information, see [Beam IO\nPerformance](https://beam.apache.org/performance/).\n\n\u003cbr /\u003e\n\nBest practices\n--------------\n\n- Avoid using [`watchForNewFiles`](https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/io/TextIO.Read.html#watchForNewFiles-org.joda.time.Duration-org.apache.beam.sdk.transforms.Watch.Growth.TerminationCondition-) with\n Cloud Storage. This approach scales poorly for large production\n pipelines, because the connector must keep a list of seen files in memory. The\n list can't be flushed from memory, which reduces the working memory of workers\n over time. Consider using\n [Pub/Sub notifications for Cloud Storage](/storage/docs/pubsub-notifications)\n instead. For more information, see\n [File processing patterns](https://beam.apache.org/documentation/patterns/file-processing/).\n\n- If both the filename and the file contents are useful data, use the\n [`FileIO`](https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/io/FileIO.html) class to read filenames. For example, a filename might\n contain metadata that is useful when processing the data in the file. For more\n information, see\n [Accessing filenames](https://beam.apache.org/documentation/patterns/file-processing/).\n The [`FileIO` documentation](https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/io/FileIO.html) also shows an example of this pattern.\n\nExample\n-------\n\nThe following example shows how to read from Cloud Storage. \n\n### Java\n\n\nTo authenticate to Dataflow, set up Application Default Credentials.\nFor more information, see\n\n[Set up authentication for a local development environment](/docs/authentication/set-up-adc-local-dev-environment).\n\n import org.apache.beam.sdk.Pipeline;\n import org.apache.beam.sdk.PipelineResult;\n import org.apache.beam.sdk.io.TextIO;\n import org.apache.beam.sdk.options.Description;\n import org.apache.beam.sdk.options.PipelineOptions;\n import org.apache.beam.sdk.options.PipelineOptionsFactory;\n import org.apache.beam.sdk.transforms.MapElements;\n import org.apache.beam.sdk.values.TypeDescriptors;\n\n public class ReadFromStorage {\n public static Pipeline createPipeline(Options options) {\n var pipeline = Pipeline.create(options);\n pipeline\n // Read from a text file.\n .apply(TextIO.read().from(\n \"gs://\" + options.getBucket() + \"/*.txt\"))\n .apply(\n MapElements.into(TypeDescriptors.strings())\n .via(\n (x -\u003e {\n System.out.println(x);\n return x;\n })));\n return pipeline;\n }\n }\n\n\u003cbr /\u003e\n\nWhat's next\n-----------\n\n- Read the [`TextIO`](https://beam.apache.org/releases/javadoc/current/org/apache/beam/sdk/io/TextIO.html) API documentation.\n- See the list of [Google-provided templates](/dataflow/docs/guides/templates/provided-templates)."]]
