Getting binary object contents is extremely slow

[Johannes Erdfelt]

When using object_store.get_object or object_store.download_object on a large binary object, it is *very* slow.

After some debugging, I traced it down to a poorly placed call to response.text. Since there is no encoding specified for the content, requests will try to automatically figure out the encoding and that is *very* slow on large binary content.

The call to response.text in openstack/transport.py:_log_response should not happen in the case of binary content.

[Everett Toews (everett-toews)]

The code causing this bug was recently completely replaced.

Can you please try this again with version 0.7.1 and see if the situation has improved?

[OpenStack Infra (hudson-openstack)]

Change abandoned by Monty Taylor (<email address hidden>) on branch: master
Review: https://review.openstack.org/247172