I am getting a bad gateway error when accessing the admin section. In the nginx log the error is "upstream sent too big header while reading response header from upstream" I have followed advice for this error of setting the fastcgi_buffers, fastcgi_buffers_size, and fastcgi_busy_buffer_size.

I attempted the 128K 256K values I found in post this did not help. I attempted to increase the sizes to 1M and 2M then 2M and 4M. Nothing I have tried has worked to get this running.

I also set proxy buffer sizes to ensure that was not the issue.

Is there something I am missing or some reason this is not working?

I have successfully installed silverstripe on many servers over the past couple of years and never ran into this problem before.

It's one of them. Client buffers perhaps.
Also 128 too small. 256 maybe, but 1M should perhaps do it.
It's caused by the http header response, SS includes custom values which fill it up. Hence it happens mostly in CMS (where it lists client side requirements, css&js)

http://forum.nginx.org/read.php?2,44091,44122#msg-44122 (see previous/op in thread)