All Fuel CI tests constantly fail twice with connect to keystone in the very beginning of the test:
2014-08-13 12:13:41,393 - INFO nailgun_client.py:30 -- Initiate Nailgun client with url http://10.108.10.2:8000
2014-08-13 12:13:41,393 - INFO http.py:26 -- Initiate HTTPClient with url http://10.108.10.2:8000
2014-08-13 12:13:41,394 - INFO http.py:37 -- Initialize keystoneclient with url http://10.108.10.2:5000/v2.0
2014-08-13 12:15:48,650 - WARNING http.py:47 -- Cant establish connection to keystone with url http://10.108.10.2:5000/v2.0
Create environment and set up master node ... 2014-08-13 12:15:49,035 - INFO nailgun_client.py:30 -- Initiate Nailgun client with url http://10.108.10.2:8000
2014-08-13 12:15:49,035 - INFO http.py:26 -- Initiate HTTPClient with url http://10.108.10.2:8000
2014-08-13 12:15:49,035 - INFO http.py:37 -- Initialize keystoneclient with url http://10.108.10.2:5000/v2.0
2014-08-13 12:17:56,266 - WARNING http.py:47 -- Cant establish connection to keystone with url http://10.108.10.2:5000/v2.0
It happens because class HTTPClient tries to authenticate keystone client during the class initialisation.
The class HTTPClient is created only when the test has just started, but environment isn't running at this moment. This is caused delays for about 5 minutes while keystone is trying to authenticate at start of every system test.
Given that the class HTTPClient cannot perform authentication of keystone when it is created but each method of this class has workaround with re-trying of keystone authentication, I propose to remove self.authenticate() from __init__ method in HTTPClient.
There will be no loss of functionality because HTTPClient still cannot perform authentication of keystone, but keystone is authenticated by any method of HTTPClient class (for example: during ntpd time syncronisation) when environment has created.
Fix proposed to branch: master
Review: https:/