将传输管理器从版本 1 迁移到版本 2 适用于 Java 的 Amazon SDK

本迁移指南涵盖了传输管理器 v1 和 S3 Transfer Manager v2 之间的主要区别，包括构造函数更改、方法映射和常见操作的代码示例。查看这些差异后，您可以成功迁移现有的 Transfer Manager 代码，以利用 v2 中改进的性能和异步操作。

概览

S3 传输管理器 v2 对传输管理器 API 进行了重大更改。S3 Transfer Manager v2 Amazon 基于异步操作构建，可提供更好的性能，尤其是在您使用基于 CRT 的 Amazon S3 客户端时。

主要区别

高级别更改

Maven 依赖项

<dependencyManagement>
    <dependencies>
        <dependency>
            <groupId>com.amazonaws</groupId>
            <artifactId>aws-java-sdk-bom</artifactId>
            <version>>1.12.7871</version>
            <type>pom</type>
            <scope>import</scope>
        </dependency>
    </dependencies>
</dependencyManagement>
<dependencies>
    <dependency>
        <groupId>com.amazonaws</groupId>
        <artifactId>aws-java-sdk-s3</artifactId>
    </dependency>
</dependencies>

<dependencyManagement>
    <dependencies>
        <dependency>
            <groupId>software.amazon.awssdk</groupId>
            <artifactId>bom</artifactId>
            <version>2.31.682</version>
            <type>pom</type>
            <scope>import</scope>
        </dependency>
    </dependencies>
</dependencyManagement>
<dependencies>
    <dependency>
        <groupId>software.amazon.awssdk</groupId>
        <artifactId>s3-transfer-manager</artifactId>
    </dependency>
    <!-- Optional: For enhanced performance with Amazon CRT -->
    <dependency>
        <groupId>software.amazon.awssdk.crt</groupId>
        <artifactId>aws-crt</artifactId>
        <version>0.38.53</version>
    </dependency>
</dependencies>

¹ 最新版本。 ² 最新版本。 ³ 最新版本。

客户端构造函数迁移

支持的构造函数（自动迁移）

不支持的构造函数（需要手动迁移）

手动迁移示例

V1 代码：

AmazonS3 s3Client = AmazonS3ClientBuilder.defaultClient();
TransferManager transferManager = new TransferManager(s3Client);

V2 代码：

// Create an `S3AsyncClient` with similar configuration
S3AsyncClient s3AsyncClient = S3AsyncClient.builder()
    .credentialsProvider(DefaultCredentialsProvider.create())
    .build();

// Provide the configured `S3AsyncClient` to the S3 transfer manager builder.
S3TransferManager transferManager = S3TransferManager.builder()
    .s3Client(s3AsyncClient)
    .build();

客户端方法迁移

目前，迁移工具支持基本copydownload、upload、uploadDirectory、downloadDirectory、resumeDownload、和resumeUpload方法。

核心传输方法

可续传方法

生命周期方法

不支持的 V1 客户端方法

copy带TransferStateChangeListener参数的方法

// V1 ----------------------------------------------------------------------------------------------
// Initialize source S3 client
AmazonS3 s3client = AmazonS3ClientBuilder.standard()
                .withRegion("us-west-2")
                .build();
                
// Initialize Transfer Manager
TransferManager tm = TransferManagerBuilder.standard()
                .withS3Client(srcS3)
                .build();

CopyObjectRequest copyObjectRequest = new CopyObjectRequest(
                "amzn-s3-demo-source-bucket",
                "source-key",         
                "amzn-s3-demo-destination-bucket", 
                "destination-key"    
        );

TransferStateChangeListener stateChangeListener = new TransferStateChangeListener() {
            @Override
            public void transferStateChanged(Transfer transfer, TransferState state) {
              //Implementation of the TransferStateChangeListener
            }
        };

Copy copy = tm.copy(copyObjectRequest, srcS3, stateChangeListener);


// V2 ----------------------------------------------------------------------------------------------
S3AsyncClient s3AsyncClient = S3AsyncClient.builder()
                .region(Region.US_WEST_2)          
                .build();

S3TransferManager transferManager = S3TransferManager.builder()
                .s3Client(s3AsyncClient)
                .build();

// Create transfer listener (equivalent to TransferStateChangeListener in v1)                                
TransferListener transferListener = new TransferListener() {
            @Override
            public void transferInitiated(Context.TransferInitiated context) {
               //Implementation
               System.out.println("Transfer initiated");
            }

            @Override
            public void bytesTransferred(Context.BytesTransferred context) {
                //Implementation
                System.out.println("Bytes transferred");
            }

            @Override
            public void transferComplete(Context.TransferComplete context) {
                //Implementation
                System.out.println("Transfer completed!");
            }

            @Override
            public void transferFailed(Context.TransferFailed context) {
                //Implementation
                System.out.println("Transfer failed");
            }
        };

CopyRequest copyRequest = CopyRequest.builder()
                              .copyObjectRequest(req -> req
                                  .sourceBucket("amzn-s3-demo-source-bucket")
                                  .sourceKey("source-key")
                                  .destinationBucket("amzn-s3-demo-destination-bucket")
                                  .destinationKey("destination-key")
                               )
                                .addTransferListener(transferListener) // Configure the transferListener into the request
                                .build();
  
Copy copy = transferManager.copy(copyRequest);  

download带S3ProgressListener参数的方法

// V1 ----------------------------------------------------------------------------------------------
S3ProgressListener progressListener = new S3ProgressListener() {
        @Override
        public void progressChanged(com.amazonaws.event.ProgressEvent progressEvent) {
            long bytes = progressEvent.getBytesTransferred();
            ProgressEventType eventType = progressEvent.getEventType();
            // Use bytes and eventType as needed
        }

        @Override
        public void onPersistableTransfer(PersistableTransfer persistableTransfer) {

        }
    };

Download download1 = tm.download(getObjectRequest, file, progressListener); 
Download download2 = tm.download(getObjectRequest, file, progressListener, timeoutMillis)
Download download3 = tm.download(getObjectRequest, file, progressListener, timeoutMillis, true)

// V2 ----------------------------------------------------------------------------------------------
TransferListener transferListener = new TransferListener() {
    @Override
    public void transferInitiated(Context.InitializedContext context) {
        // Equivalent to ProgressEventType.TRANSFER_STARTED_EVENT
        System.out.println("Transfer initiated");
    }

    @Override
    public void bytesTransferred(Context.BytesTransferred context) {
        // Equivalent to ProgressEventType.REQUEST_BYTE_TRANSFER_EVENT
        long bytes = context.bytesTransferred();
        System.out.println("Bytes transferred: " + bytes);
    }

    @Override
    public void transferComplete(Context.TransferComplete context) {
        // Equivalent to ProgressEventType.TRANSFER_COMPLETED_EVENT
        System.out.println("Transfer completed");
    }

    @Override
    public void transferFailed(Context.TransferFailed context) {
        // Equivalent to ProgressEventType.TRANSFER_FAILED_EVENT
        System.out.println("Transfer failed: " + context.exception().getMessage());
    }
};
DownloadFileRequest downloadFileRequest = 
                         DownloadFileRequest.builder()
                             .getObjectRequest(getObjectRequest)
                             .destination(file.toPath())
                             .addTransferListener(transferListener)
                             .build();

// For download1
FileDownload download = transferManager.downloadFile(downloadFileRequest);

// For download2
CompletedFileDownload completedFileDownload = download.completionFuture()
                                                  .get(timeoutMillis, TimeUnit.MILLISECONDS);

// For download3, the v2 SDK does not have a direct equiavalent to the `resumeOnRetry` method of v1.
// If a download is interrupted, you need to start a new download request.

downloadDirectory具有 4 个或更多参数的方法

// V1 ----------------------------------------------------------------------------------------------
KeyFilter filter = new KeyFilter() {
            @Override
            public boolean shouldInclude(S3ObjectSummary objectSummary) {
                //Filter implementation
            }
        };
MultipleFileDownload multipleFileDownload = tm.downloadDirectory(bucketName, keyPrefix, destinationDirectory, filter);

// V2 ----------------------------------------------------------------------------------------------
// The v2 SDK does not have a direct equiavalent to the `resumeOnRetry` method of v1.
// If a download is interrupted, you need to start a new download request.
DownloadFilter filter = new DownloadFilter() {
            @Override
            public boolean test(S3Object s3Object) {
                // Filter implementation.
            }
        };

DownloadDirectoryRequest downloadDirectoryRequest = 
                              DownloadDirectoryRequest.builder()
                                  .bucket(bucketName)
                                  .filter(filter)
                                  .listObjectsV2RequestTransformer(builder -> builder.prefix(keyPrefix))
                                  .destination(destinationDirectory.toPath())
                                  .build();
                                                                            
DirectoryDownload directoryDownload = transferManager.downloadDirectory(downloadDirectoryRequest);

upload带ObjectMetadata参数的方法

// V1 ----------------------------------------------------------------------------------------------ObjectMetadata metadata = new ObjectMetadata();
ObjectMetadata metadata = new ObjectMetadata();

metadata.setContentType("text/plain");        // System-defined metadata
metadata.setContentLength(22L);               // System-defined metadata
metadata.addUserMetadata("myKey", "myValue"); // User-defined metadata

PutObjectRequest putObjectRequest = new PutObjectRequest(bucketName, key, inputStream, metadata);

Upload upload = transferManager.upload("amzn-s3-demo-bucket", "my-key", inputStream, metadata);

// V2 ----------------------------------------------------------------------------------------------
/* When you use an InputStream to upload in V2, you should specify the content length 
   and use `RequestBody.fromInputStream()`. 
   If you don't provide the content length, the entire stream will be buffered in memory. 
   If you can't determine the content length, we recommend using the CRT-based S3 client.
*/
Map<String, String> userMetadata = new HashMap<>();
userMetadata.put("x-amz-meta-myKey", "myValue");

PutObjectRequest putObjectRequest = 
                        PutObjectRequest.builder()
                            .bucket("amzn-s3-demo-bucket1")
                            .key("k")
                            .contentType("text/plain") //System-defined metadata usually has separate methods in the builder.
                            .contentLength(22L)
                            .metadata(userMetadata) //metadata() is only for user-defined metadata.
                            .build();

UploadRequest uploadRequest = 
                        UploadRequest.builder()
                            .putObjectRequest(putObjectRequest)
                            .requestBody(AsyncRequestBody.fromInputStream(stream, 22L, executor))
                            .build();
                                                   
transferManager.upload(uploadRequest).completionFuture().join();

uploadDirectory带ObjectMetadataProvider参数

// V1 ----------------------------------------------------------------------------------------------
tm.uploadDirectory(bucketName, virtualDirectoryKeyPrefix, directory, includeSubdirectories, metadataProvider)
tm.uploadDirectory(bucketName, virtualDirectoryKeyPrefix, directory, includeSubdirectories, metadataProvider, taggingProvider)
tm.uploadDirectory(bucketName, virtualDirectoryKeyPrefix, directory, includeSubdirectories, metadataProvider, taggingProvider, cannedAclProvider)

// V2 ----------------------------------------------------------------------------------------------
UploadDirectoryRequest request = UploadDirectoryRequest.builder()
                                  .bucket(bucketName)
                                  .s3Prefix(virtualDirectoryKeyPrefix)
                                  .source(directory.toPath())
                                  .maxDepth(includeSubdirectories ? Integer.MAX_VALUE : 1)
                                  .uploadFileRequestTransformer(builder -> {
                                      // 1.Replace `ObjectMetadataProvider`, `ObjectTaggingProvider`, and `ObjectCannedAclProvider` with an
                                        // `UploadFileRequestTransformer` that can combine the functionality of all three *Provider implementations.
                                        // 2. Convert your v1 `ObjectMetadata` to v2 `PutObjectRequest` parameters.
                                        // 3. Convert your v1 `ObjectTagging` to v2 `Tagging`.
                                        // 4. Convert your v1 `CannedAccessControlList` to v2 `ObjectCannedACL`.
                                  })
                                  .build();
        
DirectoryUpload directoryUpload = transferManager.uploadDirectory(request);

模型对象迁移

在中 Amazon SDK for Java 2.x，许多TransferManager模型对象已经过重新设计，不再支持 v1 模型对象中可用的几个 getter 和 setter 方法。

在 v2 中，您可以使用CompletableFuture<T>类在传输完成时执行操作，无论是成功还是出现异常。如果需要，您可以使用该join()方法等待完成。

核心传输对象

持久性对象

结果对象

配置对象

不支持的对象

TransferManagerBuilder 配置迁移

配置更改

您需要为 v2 传输管理器设置的配置更改取决于您使用的 S3 客户端。您可以选择 Amazon 基于 CRT 的 S3 客户端或基于 Java 的标准的 S3 异步客户端。有关差异的信息，请参阅中的 S3 客户端 Amazon SDK for Java 2.x主题。

行为更改

异步操作

V1（屏蔽）：

Upload upload = transferManager.upload("amzn-s3-demo-bucket", "key", file);
upload.waitForCompletion(); // Blocks until complete

V2（异步）：

FileUpload upload = transferManager.uploadFile(UploadFileRequest.builder()
    .putObjectRequest(PutObjectRequest.builder()
        .bucket("amzn-s3-demo-bucket")
        .key("key")
        .build())
    .source(file)
    .build());

CompletedFileUpload result = upload.completionFuture().join(); // Blocks until complete
// Or handle asynchronously:
upload.completionFuture().thenAccept(result -> {
    System.out.println("Upload completed: " + result.response().eTag());
});

错误处理

V1：如果任何子请求失败，则目录传输将完全失败。

V2：即使某些子请求失败，目录传输也能成功完成。明确检查错误：

DirectoryUpload directoryUpload = transferManager.uploadDirectory(request);
CompletedDirectoryUpload result = directoryUpload.completionFuture().join();

// Check for failed transfers
if (!result.failedTransfers().isEmpty()) {
    System.out.println("Some uploads failed:");
    result.failedTransfers().forEach(failed -> 
        System.out.println("Failed: " + failed.exception().getMessage()));
}

通过字节范围提取进行并行下载

在 v2 SDK 中启用自动并行传输功能后，S3 传输管理器将使用字节范围提取来并行检索对象的特定部分（分段下载）。使用 v2 下载对象的方式不取决于对象最初的上传方式。所有下载都可以从高吞吐量和并发性中受益。

相比之下，对于 v1 的传输管理器，对象最初的上传方式确实很重要。v1 Transfer Manager 检索对象的各个部分的方式与上传分段的方式相同。如果对象最初是作为单个对象上传的，则 v1 传输管理器无法通过使用子请求来加速下载过程。