[
https://issues.apache.org/jira/browse/JCLOUDS-1487?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16743544#comment-16743544
]
Alexander Chernavin commented on JCLOUDS-1487:
----------------------------------------------
[~gaul], here is what I found so far. If content length is set to -1 or if MD5
needs to be calculated, then Azure SDK's CloudBlockBlob tries to read from the
input stream to calculate stream length and it's MD5. Then it resets the stream
and uploads it using putBlob API.
However, there is some threshold in place here. If input stream length is more
than 128Mb, then putBlob REST API is skipped and some streaming approach is
used. It's not clear yet what exactly happens during this streaming and what
kind of corresponding REST API is called during this streaming. It might be
related to [Block blobs
explanation|https://docs.microsoft.com/en-us/rest/api/storageservices/understanding-block-blobs--append-blobs--and-page-blobs#about-block-blobs]:
_Storage clients default to a 128 MB maximum single blob upload, settable using
the SingleBlobUploadThresholdInBytes property of the BlobRequestOptions object.
When a block blob upload is larger than the value in this property, storage
clients break the file into blocks._
> Blob content length is mandatory while uploading to Azure
> ---------------------------------------------------------
>
> Key: JCLOUDS-1487
> URL: https://issues.apache.org/jira/browse/JCLOUDS-1487
> Project: jclouds
> Issue Type: Bug
> Components: jclouds-blobstore
> Affects Versions: 2.1.1
> Reporter: Alexander Chernavin
> Priority: Major
> Labels: azureblob
> Attachments: length_exception.txt
>
>
> Content length is mandatory while uploading to Azure using JClouds portable
> Blob Store API
> {code:java}
> Blob blob = blobStore.blobBuilder(blobName)
> .payload(new InputStreamPayload(inputStream))
> .contentType(contentType)
> .build();
> blobStore.putBlob(container, blob);
> {code}
> [BindAzureBlobMetadataToRequest|https://github.com/fusesource/jclouds/blame/master/providers/azureblob/src/main/java/org/jclouds/azureblob/binders/BindAzureBlobMetadataToRequest.java]
> class checks that content length is specified and is greater than zero in
> lines 58-59. However, according to the Azure SDK documentation, Azure
> [CloudBlockBlob|https://azure.github.io/azure-sdk-for-java/com/microsoft/azure/storage/blob/CloudBlockBlob.html#upload-java.io.InputStream-long-]
> also accepts -1 if content length is unknown:
> {{length}} - A {{long}} which represents the length, in bytes, of the stream
> data, or -1 if unknown.
>
> We have a use case when we need to stream a big chunk of data of unknown size
> directly into a cloud storage. Portable JClouds API allows to upload files
> into OpenStack SWIFT without specifying their size, but it fails on Azure.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
