[BUG] copyFromUrlWithResponse fails with 404 CannotVerifyCopySource if blob name contains %20

[daolis]

Scanario

• Copying Container of Blobstore A to Container of Blobstore B.
• A Blob exists with the name my%20testfile.txt. You see this name in the AzurePortal (Its not the URL Encoded name)

Application functionality:

1. Get list of all Blobs from Blobstore A
2. Create a BlockBlobClient with the name from the source blob (containerClient.getBlobClient(name)
3. Add a SASToken to the BlobURL
4. Trigger server side copy process to create Blob in Blobstore B with URL from source blob

Example:

1. Get blob with name my%20testfile.txt
2. Create BlockBlobClient:
   URL is e.g. https://xyz.blob.core.windows.net/container/my testfile.txt (URL Encoding)
3. Add SAS Token
4. Trigger server side copy: -> 404, because it tries to copy the file my testfile.txt

Error
Exception: com.azure.storage.blob.models.BlobStorageException: Status code 404
Errorcode: CannotVerifyCopySource
Message: The specified blob does not exist.

Reproduce

• Create a blob with %20 in its name.
• Create a BlockBlobClient with this name
• Optional: Add a SAS Token if not public readable
• Trigger Copy Process with: targeBlockBlobClient.copyFromUrlWithResponse(...)

Expected behavior
Successful copy of the file

Workaround
.. for now is to replace the character % with %20.

1. my%20testfile.txt to my%2020testfile.txt
2. URL of BlockBlobClient: https://xyz.blob.core.windows.net/container/my%20testfile.txt
3. Copy successful

[alzimmermsft]

@gapra-msft @rickle-msft I have a suspicion that this is related to #11183

[rickle-msft]

@daolis Thank you for reporting this.

@alzimmermsft I suspsect you are right about the connection. @gapra-msft Can you be sure to test this case for blobs while you're working on files?

[gapra-msft]

Hi @daolis

I'm having a little trouble reproducing the issue you are seeing based off the steps you provided. Could you please provide a code sample so I can further investigate?

[daolis]

Hi @gapra-msft
I created a repo reproducing this issue.
https://github.com/daolis/azureBlobCopy404Example

[gapra-msft]

Hi @daolis, Thank you for your repro code.

I believe the issue is that when you call listBlobs, it returns the raw, not encoded blob name. However, the getBlobClient method expects a url encoded blob name (especially when the blob name could possibly look url encoded).

Our recommendation is for you to first url encode the blob name before you pass it in the getBlobClient method. We have a utility method that does just that if you'd like to use it. - Utility.urlEncode

Please let me know if you have further questions/concerns.

[rickle-msft]

I am going to close this issue as we have given a recommendation on how to appropriately resolve the problem and there has been no follow up.

Please feel free to reach back out to us if you need further support.