Migrate login endpoint to v1-public#15888
Conversation
2169ee6 to
90e7d8f
Compare
90e7d8f to
3fee749
Compare
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
3fee749 to
cc6e3a5
Compare
|
There seems to be an issue with the |
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
|
@aalves08 CI gates are passing. This is ready for review |
aalves08
left a comment
There was a problem hiding this comment.
I've tested these "normal" logins:
local
LDAP - OpenLDAP
OIDC - Amazon Cognito
SAML - Okta
and all worked fine.
I ran into trouble when doing a "full logout" and logging back in (SLO feature):
OIDC - Amazon Cognito
OIDC - keycloak
SAML - Okta
when I clicked to log out of the provider as well, it did log out, but logging back in, in either cases, it did not prompt me to input the credentials again. I think that's what usually happens.
No errors on the console regarding this... check screenshot above of logout URL XHR
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
There was a problem hiding this comment.
@rak-phillip with your change we got Amazon Cognito and Keycloak (both OIDC) to logout "all" just fine.
I did try Okta SAML again and I got this error:


When I click "log out from all", It does log out the user (only from Rancher, not from the auth provider) and he can log back in again, but throws this error when logging out.
If I do "only log out from Rancher", no errors thrown, all works fine.
I don't think it's UI related, but definitely something to be fixed.
Let me know if you want to move this forward anyway. I can give you the thumbs up.
No, let's do our due diligence and sync with the backend team first. We'll want to ensure that we're aligned before merging, otherwise people testing Rancher can run into undefined behavior. |
|
I get the following when I attempt to repro with Okta: |
@rak-phillip It's not clear how the repro is done. Do you pass in the the session token? Works for me just fine |
|
@aalves08 The Okta issue isn't related to this change and I can confirm that this is existing behavior present in 2.13.0. I created a follow-up issue to better investigate and address rancher/rancher#53944 |
| description: 'UI session', | ||
| responseType: 'cookie', | ||
| ...body | ||
| }, { redirectUnauthorized: false }); |
There was a problem hiding this comment.
pretty sure this is still needed, which will prevent the ui trying to go to a 404 page if the user 401 fails single-sign on?
dashboard/shell/plugins/steve/actions.js
Line 209 in 1472be4
There was a problem hiding this comment.
Indeed, you are correct. Thanks, it initially wasn't clear how redirectUnauthorized: false mapped from the previous doAction() abstraction to management/request.
Signed-off-by: Phillip Rak <rak.phillip@gmail.com>
|
@aalves08 @richard-cox I addressed the |
aalves08
left a comment
There was a problem hiding this comment.
LGTM. @richard-cox what about you?
|
This should be ready to merge based on oob discussions with @richard-cox |
Summary
This migrates the login from
v3-publictov1-public.Fixes # 15326Contributes to #15326
Occurred changes and/or fixed issues
v3-public/authProviderstov1-public/authprovidersv1-public/loginto loginTechnical notes summary
This will be easier to work with if #15613 merges first.
Areas or cases that should be tested
You can test with https://hub.docker.com/r/rancher/rancher/tags?name=e9cf13a or later and use the
v3-publicfeature flag to disable/v3-publicendpoints.We will need to validate the following:
Areas which could experience regressions
The login methods described above.
Screenshot/Video
NA
Checklist
Admin,Standard UserandUser Base