Skip to content

Uploading files that share the same base filename to S3 can overwrite each other #294

Open
@briansmith-epic

Description

@briansmith-epic

When performing an upload via the SG API, we have multiple files that share a common filename but are in completely different paths. If we try to upload both of those simultaneously, the AWS upload URL is the same for both.

In _upload_to_storage(), os.path.basename(path) is passed to the _get_attachment_upload_info call. On the AWS side, that means that no matter the parent directory, any file that shares the base name will get a duplicate upload URL.

This only occurs if the _upload_to_storage() call happens concurrently : I'm guessing that the SG backend uses a timestamp to differentiate calls, because currently we do this about 30 times, but only when multiple get called at the same time do we get a duplicate upload URL.

The culprit line is here.

Recommend changing:
filename = os.path.basename(path)
to something like:
filename = os.path.abspath(path)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions