When traffic goes out, it is up to the interfaces to decide which IP to use. Our binds only insure that it listens on the correct port. In general, outbound would use the primary IP/interface and it's up to the software to decide if you can bind differently. Generally, servers won't have those options.
That kind of issue would not be something we could fix, because it's not within the scope of what we can control. You'd need to see if docker would let you specify that IP (google for me is getting mixed results, like it's partly implemented but not perfect)
I ran into the same issue. If mailgun has the wrong values, I get random error 500's.
EDIT: I went ahead and created an issue at github. https://github.com/PufferPanel/PufferPanel/issues/660
piousminion created this issue in PufferPanel/PufferPanel
error 500 if mailgun settings are incorrect
Make sure port 5656 is accessible from the outside. You should be able to navigate to https://yourscalesIP:5656 and have it load "Scales Management Daemon". If it does not, check /srv/scales/logs for the Scales error log.
The issue is the ip binding setting of your node.
Head over to your pufferpanel with an administrator user.
Go to List Nodes, Select the Node you want and go to Allocation.
Add the external ip address of your server that you want to use and select the ports needed.
Remember that you can't delete the old one there until the servers have changed the assigned ports to the new ip route.
The reason for this is that nodejs 4 does not natively support CentOS 6. You can follow the manual installation instructions located at https://scales.readme.io/docs/manually-installing-scales but install NodeJS through building it from source (https://nodejs.org/en/download/)