This is a bug in the lib/puppet/provider/keystone_endpoint/openstack.rb. And this bug will trigger prefetch method fail when puppet runs, this will lead to adding duplicate endpoints that already exist cause prefetch method won't return actual results.
Now keystone_endpoint is managing admin/internal/public url in only one resource. The code below is triggering this bug:
```
def self.instances
names = []
list = []
endpoints.each do |current|
name = transform_name(current[:region], current[:service_name], current[:service_type])
unless names.include?(name)
names << name
endpoint = { :name => name, current[:interface].to_sym => current }
endpoints.each do |ep_osc|
if (ep_osc[:id] != current[:id]) &&
(ep_osc[:service_name] == current[:service_name]) &&
(ep_osc[:service_type] == current[:service_type]) &&
(ep_osc[:region] == current[:region])
endpoint.merge!(ep_osc[:interface].to_sym => ep_osc)
end
end
list << endpoint
end
end
list.collect do |endpoint|
new(
:name => endpoint[:name],
:ensure => :present,
:id => "#{endpoint[:admin][:id]},#{endpoint[:internal][:id]},#{endpoint[:public][:id]}",
:region => endpoint[:admin][:region],
:admin_url => endpoint[:admin][:url],
:internal_url => endpoint[:internal][:url],
:public_url => endpoint[:public][:url]
)
end
end
```
Say I have only admin and internal url of neutron, the endpoint[:public] is nil, and endpoint[:public][:id] will throw an error of undefined method `[]' for nil:NilClass. Same of endpoint[:public][:id], endpoint[:admin][:url] and endpoint[:admin][:region] is problematic. We need add some nil? check in the code above.
Secondly, even if I can fix this bug by adding some nil check. Say I do not have admin url of neutron. What I expect to happen after a puppet run is that adding a missing admin url of neutron. But the actual behavior is that puppet can only change exist admin url to a right one but not adding a new one. See code below
```
def flush
if @property_flush && @property_hash[:id]
ids = @property_hash[:id].split(',')
if @property_flush[:admin_url]
self.class.request('endpoint', 'set', [ids[0], "--url=#{resource[:admin_url]}"])
end
if @property_flush[:internal_url]
self.class.request('endpoint', 'set', [ids[1], "--url=#{resource[:internal_url]}"])
end
if @property_flush[:public_url]
self.class.request('endpoint', 'set', [ids[2], "--url=#{resource[:public_url]}"])
end
end
@property_hash = resource.to_hash
end
```
The flush method can change a wrong endpoint to a right one, but dose not support adding a missing endpoint. So we need to add some logic here to handle that when some of admin/internal/public endpoint is missing and add the missing one.
I'm not a ruby guru but I'm trying my best to fix these bugs. I've already commit a patch and tested it. Welcome for advices.
Fix proposed to branch: master
Review: https:/