I hit a road-block initially trying to reproduce this issue. I found I had to ensure that the `~/.config/openshift/.ansible/` directory was cleared away to reproduce the issue. I cleared parent directory away as well.


> # rm -fR ~/.config/openshift/
> # atomic-openshift-installer -c /root/installer.cfg.yml install
> ... set some config options ...
>
> Gathering information from hosts...
> ERROR! Unexpected Exception: [Errno 2] No such file or directory:
>     '/root/.config/openshift/.ansible/callback_facts.yaml'
> There was a problem fetching the required information.
> Please see /tmp/ansible.log for details.

Working on figuring out the exact nature of this bug now.

You ran the installer using the config file called

> /root/installer.cfg.yml

In that file what is the value of this key?

> ansible_callback_facts_yaml

Your report says that the file 

> /root/.config/openshift/.ansible/callback_facts.yaml

does not exist. Is that the same file as the "ansible_callback_facts_yaml" key?

Jia,

I've opened https://github.com/openshift/openshift-ansible/pull/2341 with a patch that I believe fixes this issue.

Please try the patch out using these commands:

> cd `mktemp -d`
> wget -r -l1 --no-parent -A.rpm --cut-dirs=5 --no-host-directories http://file.rdu.redhat.com/~tbielawa/BZ1368296/
> sudo yum localinstall *.rpm
> atomic-openshift-installer -d -c /root/installer.cfg.yml install

Please include the generated debug log with your report:

> /tmp/installer.txt

https://github.com/openshift/openshift-ansible/pull/2341

Patch merged fixes the issue as I was able to reproduce it. Now the installer will continue instead of abort for the condition where the cached facts file doesn't exist or is otherwise unreadable. The installer will continue and then recollect the missing system facts.

(In reply to Tim Bielawa from comment #3)
> Jia,
> 
> I've opened https://github.com/openshift/openshift-ansible/pull/2341 with a
> patch that I believe fixes this issue.
> 
> Please try the patch out using these commands:
> 
> > cd `mktemp -d`
> > wget -r -l1 --no-parent -A.rpm --cut-dirs=5 --no-host-directories http://file.rdu.redhat.com/~tbielawa/BZ1368296/
> > sudo yum localinstall *.rpm
> > atomic-openshift-installer -d -c /root/installer.cfg.yml install
> 
> Please include the generated debug log with your report:
> 
> > /tmp/installer.txt

Sorry for late notice and reply. I just saw your comments of this bug. I will verify it with latest errata version right now.

Many thx for your detail info about the bug.

Version:
atomic-openshift-utils-3.3.15-1.git.0.a9fd72e.el7.noarch

Verify Steps:
1.put an installer.cfg.yml in the path /root

ansible_callback_facts_yaml: /root/.config/openshift/.ansible/callback_facts.yaml
ansible_config: /usr/share/atomic-openshift-utils/ansible.cfg
ansible_inventory_path: /root/.config/openshift/hosts
ansible_log_path: /tmp/ansible.log
deployment:
  ansible_ssh_user: root
  hosts:
  ...
  ...
  master_routingconfig_subdomain: ''
  proxy_exclude_hosts: ''
  proxy_http: ''
  proxy_https: ''
  roles:
    etcd: {}
    master: {}
    master_lb: {}
    node: {}
    storage: {}
variant: openshift-enterprise
variant_version: '3.3'
version: v2

2.run "atomic-openshift-installer -u -c /root/installer.cfg.yml install"

Verify Result:
1. Though there are no file(callback_facts.yaml) in  /root/.config/openshift/.ansible/ configured in the installer config, it will continue to collect info and create the file in the same directory as the installer.cfg.yml which located in /root for this case.
2. After gathering information, it will inform user about the hosts' location:
Wrote atomic-openshift-installer config: /root/installer.cfg.yml
Wrote Ansible inventory: /root/hosts

then,it will continue to install with ansible playbook.

btw, i checked another two cases accoding to the fix.
(1) if there is wrong file in the /root/.config/openshift/.ansible/, it will also recollect info and inform user to delete wrong callback_facts.yaml.
(2) if there is right file in the /root/.config/openshift/.ansible/, it will update the file with new collected info and then create hosts in /root(the same as installer.cfg.yml)

I think it now fixed perfectly.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2016:1933