Amazon S3 Multipart Uploads

With a single PutObject operation, you can upload objects up to 5 GB in size. However, by using the multipart uploads (e.g., CreateMultipartUpload, UploadPart, CompleteMultipartUpload, AbortMultipartUpload), you can upload object up to 5 TB in size.

Multipart uploads are designed to improve the upload experience for larger objects. With it you can upload objects in parts that can be uploaded independently, in any order, and in parallel. You can use a multipart upload for objects from 5 MB to 5 TB in size.

Amazon S3 customers are encouraged to use multipart uploads for objects greater than 100 MB.

The MultipartUploader object

The SDK has a special MultipartUploader object to make the multipart upload process as easy as possible.

use Aws\S3\MultipartUploader;
use Aws\Exception\MultipartUploadException;

$uploader = new MultipartUploader($s3Client, '/path/to/large/file.zip', [
    'bucket' => 'your-bucket',
    'key'    => 'my-file.zip',
]);

try {
    $result = $uploader->upload();
    echo "Upload complete: {$result['ObjectURL']}\n";
} catch (MultipartUploadException $e) {
    echo $e->getMessage() . "\n";
}

The uploader creates a generator of part data based on the provided source and configuration, and attempts to upload all parts. If some part uploads fail, the uploader keeps track of it and continues to upload later parts until the entire source data has been read. It then either completes the upload or throws an exception that contains information about the parts that failed to be uploaded.

Customizing a multipart upload

Custom options can be set on the CreateMultipartUpload, UploadPart, and CompleteMultipartUpload operations executed by the multipart uploader via callbacks passed to its constructor.

$source = '/path/to/large/file.zip';
$uploader = new MultipartUploader($s3Client, $source, [
    'bucket' => 'your-bucket',
    'key'    => 'my-file.zip',
    'before_initiate' => function (\Aws\Command $command) {
        // $command is a CreateMultipartUpload operation
        $command['CacheControl'] = 'max-age=3600';
    },
    'before_upload' => function (\Aws\Command $command) {
       // $command is an UploadPart operation
       $command['RequestPayer'] = 'requester';
    },
    'before_complete' => function (\Aws\Command $command) {
       // $command is a CompleteMultipartUpload operation
       $command['RequestPayer'] = 'requester';
    },
]);

Recovering from errors

When an error occurs during the multipart upload process, a MultipartUploadException is thrown. This exception provides access to the UploadState object, which contains information about the multipart upload's progress. The UploadState can be used to resume an upload that failed to complete.

$source = '/path/to/large/file.zip';
$uploader = new MultipartUploader($s3Client, $source, [
    'bucket' => 'your-bucket',
    'key'    => 'my-file.zip',
]);

do {
    try {
        $result = $uploader->upload();
    } catch (MultipartUploadException $e) {
        $uploader = new MultipartUploader($s3Client, $source, [
            'state' => $e->getState(),
        ]);
    }
} while (!isset($result));

Resuming an upload from an UploadState will only attempt to upload parts that are not already uploaded. The state object keeps track of missing parts, even if they are not consecutive. The uploader will read/seek through the provided source file to the byte ranges belonging to the parts that still need to be uploaded.

UploadState objects are serializable, so it's also possible to resume an upload in a different process. You can also get the UploadState object even when you are not handling an exception by calling $uploader->getState().

Important

Streams passed in as a source to a MultipartUploader will not be automatically rewound before uploading. If you are using a stream instead of a file path in a loop similar to the above example, you will need to reset the $source variable inside of the catch block.

$source = fopen('/path/to/large/file.zip', 'rb');
$uploader = new MultipartUploader($s3Client, $source, [
    'bucket' => 'your-bucket',
    'key'    => 'my-file.zip',
]);

do {
    try {
        $result = $uploader->upload();
    } catch (MultipartUploadException $e) {
        rewind($source);
        $uploader = new MultipartUploader($s3Client, $source, [
            'state' => $e->getState(),
        ]);
    }
} while (!isset($result));

Aborting a multipart upload

Sometimes, you may not want to resume an upload though, and would rather just abort the the whole thing when an error occurs. This is also easy using the data contained in the UploadState object.

try {
    $result = $uploader->upload();
} catch (MultipartUploadException $e) {
    // State contains the "Bucket", "Key", and "UploadId"
    $params = $e->getState()->getId();
    $result = $s3Client->abortMultipartUpload($params);
}

Asynchronous multipart uploads

Calling upload() on the MultipartUploader is a blocking request. If you are working in an asynchronous context, you can get a Promise for the multipart upload.

$source = '/path/to/large/file.zip';
$uploader = new MultipartUploader($s3Client, $source, [
    'bucket' => 'your-bucket',
    'key'    => 'my-file.zip',
]);

$promise = $uploader->promise();

Configuration

The MultipartUploader object constructor accepts the following arguments:

$client

The Aws\ClientInterface object to use for performing the transfers. This should be an instance of Aws\S3\S3Client.

$source

The source data being uploaded. This can be a path or URL to a (e.g., /path/to/file.jpg), a resource handle (e.g., fopen('/path/to/file.jpg', 'r)), or an instance of a PSR-7 stream

$config

An associative array of configuration options for the multipart upload.

The following configuration options are valid:

acl

(string) ACL to set on the object being upload. Objects are private by default.

before_complete

(callable) Callback to invoke before the CompleteMultipartUpload operation. The callback should have a function signature like function (Aws\Command $command) {...}.

before_initiate

(callable) Callback to invoke before the CreateMultipartUpload operation. The callback should have a function signature like function (Aws\Command $command) {...}.

before_upload

(callable) Callback to invoke before any UploadPart operations. The callback should have a function signature like function (Aws\Command $command) {...}.

bucket

(string, required) Name of the bucket to which the object is being uploaded.

concurrency

(int, default: int(5)) Maximum number of concurrent UploadPart operations allowed during the multipart upload.

key

(string, required) Key to use for the object being uploaded.

part_size

(int, default: int(5242880)) Part size, in bytes, to use when doing a multipart upload. This must between 5 MB and 5 GB, inclusive.

state

(Aws\Multipart\UploadState) An object that represents the state of the multipart upload and that is used to resume a previous upload. When this option is provided, the bucket, key, and part_size options are ignored.

Multipart Copies

The SDK also includes a MultipartCopy object that is used in a similar manner to the MultipartUploader but is designed for copying objects between 5GB and 5TB in size within S3.

use Aws\S3\MultipartCopy;
use Aws\Exception\MultipartUploadException;

$copier = new MultipartCopy($s3Client, '/bucket/key?versionId=foo', [
    'bucket' => 'your-bucket',
    'key'    => 'my-file.zip',
]);

try {
    $result = $copier->copy();
    echo "Copy complete: {$result['ObjectURL']}\n";
} catch (MultipartUploadException $e) {
    echo $e->getMessage() . "\n";
}