Support gh gei migrate-repo with archives stored in the local machine (--git-archive-path, --metadata-archive-path)

[timrogers]

When you run gh gei migrate-repo, the migration runs end-to-end, exporting data from your source GitHub instance and importing it into your destination GitHub instance.

In some scenarios, you may already have the Git archive and metadata archive on your local machine, and want to import it.

This is supported - but you have to upload the file to a blob storage provider yourself, generate a URL for the file and then call migrate-repo with the --git-archive-url and --metadata-archive-url arguments.

The process of importing an already-existing archive would be easier if you could pass a --git-archive-path and --metadata-archive-path instead, and ask the CLI to upload it to blob storage for you and generate a URL.

[robandpdx]

@timrogers Some customers have challenges when it comes to providing the blob storage needed for GEI migrations. With an option to use a local migration archive, perhaps it would be possible to bypass blob storage entirely. Or use a GH managed blob storage, which would provide the same experience as bypassing blob storage. Perhaps there would need to be a feature flag to enable such an option only for customers that need it, to reduce possibility of abuse.

[timrogers]

@timrogers Some customers have challenges when it comes to providing the blob storage needed for GEI migrations. With an option to use a local migration archive, perhaps it would be possible to bypass blob storage entirely. Or use a GH managed blob storage, which would provide the same experience as bypassing blob storage. Perhaps there would need to be a feature flag to enable such an option only for customers that need it, to reduce possibility of abuse.

Being able to avoid the need for blob storage entirely is definitely something on my mind, and in my opinion, something we'll need to do to support customers who don't use public cloud providers.

When I wrote this issue, the scope I had in mind was "pass a path and it is uploaded for you", but I'm happy to create a separate issue around avoiding blob storage entirely.

[timrogers]

Issue created here: #1198