Havana COI doesn't configure Heat correctly

[Juergen Brendel]

There's a glitch in the COI (Havana). The Heat config file (/etc/heat/heat.conf on the control node) contains a number of "url" parameters of type "heat_*_url":

    heat_metadata_server_url=http://127.0.0.1:8000
    heat_waitcondition_server_url=http://127.0.0.1:8000/v1/waitcondition
    heat_watch_server_url=http://127.0.0.1:8003

These URLs are given to the VM guests when they are built and when their startup/init scripts are running. In the VMs - naturally - there is no Heat server running that could be reached on the localhost (127.0.0.1) address.

Those parameters should instead use the IP address of the controller node of the OpenStack installation. For example, in my particular case, it would look like this:

    heat_metadata_server_url=http://10.0.101.21:8000
    heat_waitcondition_server_url=http://10.0.101.21:8000/v1/waitcondition
    heat_watch_server_url=http://10.0.101.21:8003

During the COI installation, the heat config file should be modified with the controller IP addresses.

[Mark T. Voelker (mvoelker)]

1.) What scenario were you using when you saw this (e.g. all_in_one, 2_role, etc)?

2.) What happens if you add this to your user.common.yaml (or a host override file)?

heat::engine::heat_metadata_server_url: 'http://your_controller_address:8000'

[Juergen Brendel (jbrendel)]

The scenario was "2_role".

Unfortunately, I'm currently not able to redo my OpenStack install, since I need it for other work. Therefore, I'm afraid I can't tell you what happens when I add it to the user.common.yaml file.

[Chris Ricker (chris-ricker)]

https://review.openstack.org/#/c/66985/

upstream PR to fix

[Chris Ricker (chris-ricker)]

Upstream fixes merged