So this is not a regression and there are two possible solutions to this issue. 

1. Properly document the 'disable_auto_scaling' marker, which should say that the gear with the ha_proxy needs to be restarted to pick-up the changes. Also would add an banner to the start script(when it detects the marker) which would say that the gear needs to be restarted, together with the command for the user.

2. Restart the ha_proxy every-time when the 'disable_auto_scaling' marker is present.

I would rather go with the first solution, cause this is not a regression and there is no point to restart the ha-proxy on every push, then the proxy doesn't make any sense.

I agree, it doesn't make sense to restart haproxy at every push.

Agree, let's go with your solution #1.

Commit pushed to master at https://github.com/openshift/origin-server

https://github.com/openshift/origin-server/commit/e8dd080bd9ba5273fbeef2c88e1b230c13b0f604
BUG 1171066: disable_auto_scaling marker requires restart of haproxy cartridge

Fix by properly documenting the 'disable_auto_scaling' marker and adding banner which tells user(when it detects the marker) upon deployment, that he needs to restart the gear or the whole application, in order to HAproxy can pick up the changes.

Confirmed.The banner has been added. When restart the Haproxy or the whole app, the 'disable_auto_scaling' marker working well.

ON devenv_5381.