File Transfer Between Cloud Services (AWS to Azure)

Sometimes is a requirement to transfer files from one cloud service to another. At least that was the challenge that I encountered while developing a file classification web service. Mid development cycle management changed cloud provider because of some juicy Azure discounts and the all so great office 365 integration that everyone loves.

Because of this, the compromise reached to be able to continue at the same development speed and still lower the cloud services bill, was to continue to work on AWS in the web service, but the files, after finished the process, must be transferred to Azure.

With all the restrictions regarding failure upon transfer and other things, there are two that are the most relevant for this particular case, which are:

After much thinking and going back and forward with the team, got the idea to request the file through a stream of data, and then upload it chunk by chunk into the cloud. In that way the only part of the file that we will have in our harddrive is the current chunk of the file that we are transferring but nothing more.

The requirements to implement this code, are the packages and the Python version which I was working with at the time. Also is expected a basic knowledge of AWS S3 and Azure APIs.

  boto3 = "==1.21"
  azure-storage-blob = "==12.10"
  requests = "==2.27"

  python_version = "3.9"

For azure, there is a CLI tool called azcopy, from Microsoft that allows for multithreaded parallel download and upload of files, and given the restrictions of your own problem is worth checking out. In our case, the tool still needs to have the file in the hard-drive, which as we said before is something we cannot have.

As a start we need the azure blob client for the container and the destination file that we are going to create.

account_url = (

blob_client = BlobServiceClient(

Now for the AWS S3 client we have:

aws_session = boto3.session(

s3_client = aws_session.client('s3')

In case the files are private, you need to take into account that we have to generate a public URL. In AWS the maximum amount of time that a pre signed url of the object has is 7 days, which are 604800 seconds. If the file can’t be downloaded and transferred in the 7 days, the file will remain as a partial transfer. In our case the biggest file needed a 4 day continuous stream to fully transfer.

object_url = s3_client.generate_presigned_url(
        "Bucket": aws_storage_bucket_name,
        "key": aws_object_key,

Then we have to create the stream from the file that we are requesting, and upload it to the Azure blob client.

When we request a file as a stream, what is returned is an iterable object that the blob uploader takes and upload one chunk at a time in an ordered manner.

object_stream = requests.get(object_url, stream=True)

And with this solution we were able to seamlessly transfer all the files from one provider to another, without slowing down our work, and still benefiting with the lower bill for storage.

Tags: python, AWS, Azure