| 2021-06-21 08:55:28 |
Mark Goddard |
description |
Hi,
I found strange behaviour when openstack was deployed to two regions, openid keystone federation with public endpoints defined for both regions (edge architecture) - RegionOne, RegionTwo.
Horizon also available in both regions.
Unfortunately I don't have test stack right now, but I will try to explain behaviour what I've seen.
I noticed below pattern of behavior when using horizon and switching between RegionOne and RegionTwo (for example list instances for RegionOne, switch to RegionTwo and vice versa):
From horizon RegionOne operations against RegionOne always working.
From horizon RegionOne operations against RegionTwo are broken from time to time.
From horizon RegionTwo operations against RegionOne always working.
From horizon RegionTwo operations against RegionTwo are broken from time to time.
The above pattern of behavior made me look at the configuration of RegionTwo and found that kolla is not setting region_name in keystone middleware section.
When I've set region_name = RegionOne in regionOne and region_name = RegionTwo in regiontwo, everything started to working flawlessly.
From keystone middleware opts.py -
cfg.StrOpt('region_name',
help='The region in which the identity server can be found.'), |
Hi,
I found strange behaviour when openstack was deployed to two regions, openid keystone federation with public keystone endpoints defined for both regions (edge architecture) - RegionOne, RegionTwo.
Horizon also available in both regions.
Unfortunately I don't have test stack right now, but I will try to explain behaviour what I've seen.
I noticed below pattern of behavior when using horizon and switching between RegionOne and RegionTwo (for example list instances for RegionOne, switch to RegionTwo and vice versa):
From horizon RegionOne operations against RegionOne always working.
From horizon RegionOne operations against RegionTwo are broken from time to time.
From horizon RegionTwo operations against RegionOne always working.
From horizon RegionTwo operations against RegionTwo are broken from time to time.
The above pattern of behavior made me look at the configuration of RegionTwo and found that kolla is not setting region_name in keystone middleware section.
When I've set region_name = RegionOne in regionOne and region_name = RegionTwo in regiontwo, everything started to working flawlessly.
From keystone middleware opts.py -
cfg.StrOpt('region_name',
help='The region in which the identity server can be found.'), |
|
