DracoBlue

Hey @jhaals! What do you think about those changes? For our deployment we need to be able to disable the "one-click" link - thus this PR would also fix me...

That's a good idea!

If you increase `proxy_max_temp_file` or disable `proxy_buffering` with off, it should work. I recently blogged about this issue at https://dracoblue.net/dev/hanging-subrequests-in-nginx/

trickle works only for non-forking processes. if you try with e.g. curl it works :)

``` FROM alpine:3.9 RUN apk add --update autoconf automake libtool vim bash alpine-sdk libevent-dev libevent libtirpc-dev RUN cd /tmp/ \ && wget https://github.com/mariusae/trickle/archive/09a1d955c6554eb7e625c99bf96b2d99ec7db3dc.zip \ && unzip *.zip \ && cd...

I am interested in this, too.

We fixed it in our setup with an environment variable for the socks proxy like this: ```text export ALL_PROXY=socks5://proxy.example.org:1080 ```

@srikiz / @longwuyuan / @shaxiaozz Any way to add this breaking change to https://github.com/kubernetes/ingress-nginx/releases/tag/controller-v1.12.0 ? That `annotations-risk-level: Critical` is now `annotations-risk-level: High`. The change made most of our ingresses not...