Honor chef_server_url for sandbox uploads

[sbbowers]

Description

Chef uploads cookbooks transactionally by generating a unique sandbox
URL to upload all the associated files to using a "POST
/organizations//sandboxes" operation.

This POST operation returns the URL of the sandbox that should then be
used to PUT the cookbook's files to. This URL is pre-resolved, based on
the configuration of the chef server, which can be different from what
is configured in the client.

In special cases of network configuration, this discrepancy forces the
client to use a URL with a non-resolvable or unreachable host for
uploading cookbooks, rather than the scheme/host/port configured with
chef_server_url.

This change parses the sandbox URL returned by the server "POST
/organizations//sandboxes" operation, and rewrites
it honor the chef_server_url configuration for subsequent cookbook PUT operations.

Issues Resolved

• Does not resolve, but is semi-related to Impossible to upload cookbooks if non-standard port is used chef-server#50

Check List

• New functionality includes tests
• All tests pass
• RELEASE_NOTES.md has been updated if required (not required for bugfixes, required for API changes)
• All commits have been signed-off for the Developer Certificate of Origin. See https://github.com/chef/chef/blob/master/CONTRIBUTING.md#developer-certification-of-origin-dco

[stevendanna]

Thank you for the detailed description of the problem you are trying to solve. Unfortunately, this change would break cookbook uploads for anyone using Hosted Chef or anyone using their own Chef Server with cookbook storage configured to point at their own S3 bucket.

I realize this is easier to fix here, but I think we need to dig into why chef/chef-server#50 regressed and/or wasn't ever fixed.

[sbbowers]

Ah. I understand the use case with the cookbook storage back-end; thank you for pointing that out.

Would a pull request for a new setting be acceptable? It could be something like "cookbook_upload_proxy", that we could configure in our client.rb/knife.rb to allow us to over-ride the host/port of the sandbox URLs provided from the server. This would also ensure continuing compatibility for the various storage back-ends by default.

I agree that this problem is ultimately a deficiency in the server code, however, the server can't (and probably shouldn't) know about every potential client configuration used to connect to it. A cookbook proxy setting on the client side could also be used to create compatibility with various chef server configurations, as the API itself would have to change in order to fix chef/chef-server#50.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.