On 07/22/2017 07:14 AM, Lance Bragstad wrote:
After a little head scratching and a Pantera playlist later, we ended up
figuring out the main causes. The failures can be found in the gate [0].
The two failures are detailed below:
1.) Keystoneauth version 3.0.0 added a lot of functionality and might
return a different url depending on discovery. Keystonemiddleware use to
be able to mock urls to keystone in this case because keystoneauth
didn't modify the url in between. Keystonemiddleware didn't know how to
deal with the new url and the result was a Mock failure. This is
something that we can fix in keystonemiddleware once we have a version
of keystoneauth that covers all discovery cases and does the right
thing. NOTE: If you're mocking requests to keystone and using
keystoneauth somewhere in your project's tests, you'll have to deal with
this. More on that below.
Upon further digging - this one is actually quite a bit easier. There
are cases where keystoneauth finds an unversioned discovery endpoint
from a versioned endpoint in the catalog. It's done for quite a while,
so the behavior isn't new. HOWEVER - a bug snuck in that caused the url
it infers to come back without a trailing '/'. So the requests_mock
entry in keystonemiddleware was for http://keystone.url/admin/ and
keystoneauth was doing a get on http://keystone.url/admin.
It's a behavior change and a bug, so we're working up a fix for it. The
short story is though that once we fix it it should not cause anyone to
need to update requests_mock entries.
2.) The other set of failures were because keystoneauth wasn't expecting
a URL without a path [1], causing an index error. I tested the fix [2]
against keystonemiddleware and it seems to take care of the issue. Eric
is working on a fix. Once that patch is fully tested and vetted we'll
roll another keystoneauth release (3.0.1) and use that to test
keystonemiddleware to handle the mocking issues described in #1. From
there we should be able to safely bump the minimum version to 3.0.1, and
avoid 3.0.0 all together.
Patch is up for this one, and we've confirmed it fixes this issue.
Let me know if you see anything else suspicious with respect to
keystoneauth. Thanks!
[0]
http://logs.openstack.org/84/486184/1/check/gate-keystonemiddleware-python27-ubuntu-xenial/7c079da/testr_results.html.gz
[1]
https://github.com/openstack/keystoneauth/blob/5715035f42780d8979d458e9f7e3c625962b2749/keystoneauth1/discover.py#L947
[2] https://review.openstack.org/#/c/486231/1
On 07/21/2017 04:43 PM, Lance Bragstad wrote:
The patch to blacklist version 3.0.0 is working through the moment [0].
We also have a WIP patch proposed to handled the cases exposed by
keystonemiddleware [1].
[0] https://review.openstack.org/#/c/486223/
[1] https://review.openstack.org/#/c/486231/
On 07/21/2017 03:58 PM, Lance Bragstad wrote:
We have a patch up to blacklist version 3.0.0 from global-requirements
[0]. We're also going to hold bumping the minimum version of
keystoneauth until we have things back to normal [1].
[0] https://review.openstack.org/#/c/486223/
[1] https://review.openstack.org/#/c/486160/1
On 07/21/2017 03:00 PM, Lance Bragstad wrote:
I started noticing some trivial changes failing in the
keystonemiddleware gate [0]. The failures are in tests that use the
keystoneauth1 library (8 tests are failing by my count), which we
released a new version of yesterday [1]. I've proposed a patch to
blacklist keystoneauth1 3.0.0 from keystonemiddleware until we can
figure out what happened [2]. Status is being tracked in a bug against
keystonemiddleware [3], but might need to be broadened if these changes
are affecting other projects.
I'll be in -keystone working through some of the issues if you need me.
Thanks,
Lance
[0] https://review.openstack.org/#/c/486184/
[1] http://lists.openstack.org/pipermail/openstack-dev/2017-July/119969.html
[2] https://review.openstack.org/#/c/486213/
[3] https://bugs.launchpad.net/keystonemiddleware/+bug/1705770
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
