Description
What is the problem?
In AWS we use multipart uploads to upload backup files into S3. We track the multipart upload ID during this so that if we crash or in some other way orphan the upload we can abort it later on, in order to not take up space in S3 (you get charged for parts in a multipart upload that haven't been completed).
For GCP they do not expose this, so we have to upload multiple objects and then join them together to get the same functionality. We do not clean these up however.
What is the fix?
Find a way to clean up these temporary objects if the upload is interrupted