I have used az storage copy to copy a VHD page blob between storage accounts (same region and subscription). The VHD has a nominal size of 127 GiB. The 'used capacity' metric for the source account shows that the blob uses about 6 GiB whereas the same metric for the destination account shows it using the full 127 GiB. The calculation described in https://docs.microsoft.com/en-us/azure/storage/scripts/storage-blobs-container-calculate-billing-size-powershell tells the same story.

It seems that the sparseness of the blobs is not preserved by the copy process. A similar issue was noted for az storage blob download (#5872) and has since been fixed (Azure/azure-sdk-for-python#7555).

A comment on a similar issue for Azcopy (Azure/azure-storage-azcopy#391 (comment)) notes that this issue (loss of sparseness) did not happen when using az storage blob start-batch to perform the copy. From the follow-up in Azure/azure-storage-net-data-movement#194 it appears that there are at least two different back-end copy methods, one sync and one async, and that the async method preserves sparseness while (presumably) the sync method does not.

I would like to use az storage copy because it is synchronous, but I would like it to preserve sparseness.

In the meantime, I will investigate using az storage blob copy start and a wait loop around az storage blob show.

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: b8c4e415-fc8d-26d5-3213-da937fd6932c
• Version Independent ID: dcb2f2d6-fd80-495d-8582-8be859d68792
• Content: az storage
• Content Source: src/azure-cli/azure/cli/command_modules/storage/_help.py
• Service: storage
• GitHub Login: @rloutlaw
• Microsoft Alias: routlaw