Listen tcp: missing port in address /var/opt/gitlab/gitlab-workhorse/socket

I have an installation of gitlab running on centos 7.
I have an apache server with sites setup already. I had gitlab running via the standard install but I want it to proxy through apache. I followed this guide:
https://docs.gitlab.com/omnibus/settings/nginx.html#using-a-non-bundled-web-server

It looks to be working in that my apache virtualhosts are up and gitlab shows a 503 error message.

The problem is with this error:

2016-12-27_20:48:49.32715 2016/12/27 15:48:49 Starting gitlab-workhorse v1.2.1-20161222.170947
2016-12-27_20:48:49.32749 2016/12/27 15:48:49 listen tcp: missing port in address /var/opt/gitlab/gitlab-workhorse/socket

status shows:

down: gitlab-workhorse: 0s, normally up, want up; run: log: (pid 147) 30014s
run: logrotate: (pid 20432) 25s; run: log: (pid 142) 30014s
run: postgresql: (pid 20440) 25s; run: log: (pid 140) 30014s
run: redis: (pid 20449) 24s; run: log: (pid 141) 30014s
run: sidekiq: (pid 20474) 22s; run: log: (pid 151) 30014s
run: unicorn: (pid 20497) 20s; run: log: (pid 144) 30014s

gitlab-ctl reconfigure doesn’t give any error

I’ve tried many different users in the .rb

#web_server['external_users'] = ['apache']
#web_server['username'] = 'apache'
#web_server['group'] = 'apache'
#web_server['uid'] = 49
#web_server['gid'] = 49
#web_server['shell'] = '/bin/nologin'
#web_server['home'] = '/var/www'

This would happen if you changed/added gitlab_workhorse['listen_network'] = "tcp" but did not change (or add) the value of gitlab_workhorse['listen_addr']. Nginx can use a Unix Domain Socket to communicate with the Gitlab worker, but Apache cannot, so you have to change it to listen on a TCP port.

If you need more help, please show us the contents of your gitlab.rb config file.