Updating authenticators from latest in Tiled#81
Conversation
|
Wow, thanks @davidpcls! |
|
Thank you. I am going to try to fix the unit tests first, I may push a few commits. |
|
If either of you intend to commit more work to this PR, put it in Draft mode. Change the mode back when you are ready for final review. |
579d82a to
f54ef36
Compare
|
I fixed the unit tests. I also rebased the branch to main, so I had to force push the changes. |
|
Thanks for working on that @dmgav! I just got things setup so I can test with this against MS Entra, so I'll go about verifying there are no other changes required for this to work properly. Took a little longer than expected to get the Entra stuff worked out. |
This is working okay, although it doens't really work smoothly for the API based login and the http command based login isn't great, as it requires the user to copy and past token around. Compared to ldap which just logs the user in. So still some work to do here to smooth out the user experience.
This solves the problem that what was implemented was actually authenticating the application and not the user like expected. It worked but it required that the user input a code. This solves that problem so that when you click the login link, if you are already logged in with you SSO provider you'll just automatically log in to the HTTP Server. Likewise if you use the bluesky queueserver api, when you call RM.Login you'll automatically be logged in, no user interaction required.
These should correct some of the problems in the last CI workflow. I moved the LDAP and docker image into the continuous_integration folder so it matches tiled.
This addresses documentation problems, the levels were incorrect as I did not understand what the next level should have been in the docs. I've also updated the usage documentation a little to be more useful.
These allow for running the unit tests in a containerized system just like how they are done in the ci pipeline, but locally and in a way that can maximize processor usage and minimize runtime.
Testing with ms entra
|
I've added in these changes, I'm going to work on cleaning up failing unit tests now and then I will update the pull request description to match your style and provide information on testing. After that I will remove the WIP status. |
|
Please also see the related PR for the API: bluesky/bluesky-queueserver-api#62 |
This is a set of test changes intended to improve the reliability of unit testing, as the current unit tests are randomly failing due to test design. Primarily this appears to be centered around LDAP. So this work was to: * Fix for ldap errors * Hardening unit tests so they fail less frequency * Try to handle console output more reliably
|
Sorry, got excited about the unit tests passing. I need to do one last integration test first, which is why I converted it back into a WIP |
|
@dmgav , I've now tested this on remote machines and with httpserver running locally and both work. Let me know if you want an example. I'm not adding a whole lot here for evidence just because I'm not entirely sure what could be leaking security-sensitive information out. |
|
@davidpcls With Tiled v0.2.10 (released this morning) we addressed several issues in OIDC authentication. I think it would make sense to incorporate those fixes in this PR, so that QS will be fully caught up. Do you have capacity to take that on? See the diff in |
|
@danielballan, I'll take a look and see if it's something I can do. The unit tests are usually the hard part here as they take a long time to run, if anyone wants to tackle improving that it would help immensely. |
|
New Issues (6)Checkmarx found the following issues in this Pull Request
Fixed Issues (9)Great job! The following issues were fixed in this Pull Request
Communicate with Checkmarx by submitting a PR comment with @Checkmarx followed by one of the supported commands. Learn about the supported commands here. |
|
@danielballan, I think this is going to take a little bit, as I won't be able to work on this for a couple weeks and I've not made huge progress so far. If you need to get this going I'd suggest someone else take over integrating in the latest from Tiled. If not I'll work on it later. |
|
I'm looking at this right now, trying to move to a common auth package instead so we can limit the amount of churn here. Should also make unit tests easier to write. |
|
I am still working on this. I had stared on moving to a common auth package but that ballooned the changes so much that it didn't make sense to me. Instead I'm just completing this task and making anything I do match Tiled as much as possible. Later tasks I think should involve making Bluesky-HTTPServer and Tiled look more similar. At that point it'll make much, much more sense to have a common auth. You can see the updates to this PR in my own code base here: https://github.com/davidpcls/bluesky-httpserver/pull/9/changes Still a few things to clean up but I think it's looking much better. Also resolved that bug where the unit tests just fail randomly because of bad ports. |
| volumes: | ||
| - 'openldap_data:/var/lib/ldap' | ||
|
|
||
| volumes: | ||
| openldap_data: | ||
| driver: local |
There was a problem hiding this comment.
| volumes: | |
| - 'openldap_data:/var/lib/ldap' | |
| volumes: | |
| openldap_data: | |
| driver: local | |
| volumes: | |
| - 'openldap_data:/var/lib/ldap' | |
| - 'openldap_config:/etc/ldap/slapd.d' | |
| volumes: | |
| openldap_data: | |
| driver: local | |
| openldap_config: | |
| driver: local |
The container failed to start properly without this.
There was a problem hiding this comment.
How were you running this? I'm running it locally with the local docker script:
./scripts/run_ci_docker_parallel.sh --python-versions 3.11 --workers 20 --chunks 100
Which, as mentioned, is a mess but useful I think until we can speed up the tests other ways.
There was a problem hiding this comment.
It is still good to have the ability to manually start the container and run individual tests by hand. The issue that I observe is that if I try to restart the LDAP container by hand, it fails to start because /var/lib/ldap volume still exists but /etc/ldap/slapd.d volume is missing. I have to delete the volumes explicitly by running
docker compose -f continuous_integration/docker-configs/ldap-docker-compose.yml down -v
Adding openldap_config volume seems to fix the issue. It also works if openldap_data is removed from the configuration, but I am not sure if this is the right solution.
I start the LDAP container by running
bash continuous_integration/scripts/start_LDAP.sh
|
@davidpcls I am trying to authentication manually using Google provider and authentication fails. Did you test it with Google? Which provider did you use? |
|
@dmgav are you looking at the changes from https://github.com/davidpcls/bluesky-httpserver/pull/9/changes? That's where the current work is and I'm hoping to merge that in soon to this branch, then clean this branch up as I realize it wasn't as ready for review as I thought it was. I haven't tested this with Google auth yet, I've been focusing solely on entra. |
|
@danielballan This specifically is the change I'm proposing to both Tiled and bluesky-httpserver's entra authenticator for entra support: Adding in an optional parameter to check against graphAPI when requested. By default it would do just what you originally wrote and should be backwards compatible. From what I've read and my testing here at the CLS, many claims are optional, configurable, and can vary. For us none could be used as an immutable human readable username. This is partially because of our infrastructure choices and partially because MS Entra itself (and OIDC auth in general) has no concept of an enforced user profile in the auth flow, it only authenticates that an identity is known to a identity provider and provides a unique id for that user: the OID. With entra you then make a MS GraphAPI request to get profile information you require, like the system-wide username you use. |
* Updating authenticators * Migrating changes from 0.2.12 tiled in * Adding new port tests * Removing lazy import * Removing my dumb changes * Cleaning up the shutdown changes * Adding auto-upgrade ability * Fixing typo * Code review adjustments * Removing timezone changes to reduce scope * Removing unnecessary formatting changes This just makes the diff easier to understand * Some more cleanup * Fixing linting issue reported by isort * Cleaning up pre-commit check * Fixed up unit tests * Fixing linting issue * Making tests more robust * Pre-commit checks * Fixing up test * Refactoring unit tests * Fixed up unit tests after refactoring Moved things around to look more like tiled * Fixing pre-commit errors * Cleanup of the get_current_principal func It was just too large, so moving it towards tiled style as much as possible. Still not perfect but at least better. Tests need updating. * Updating unit test * Working version with MS graph API * Cleanup from pre-commit * Fixes and cleanup to the authenticators * Pre-commit fixes * Cleaning up issues with function renaming in tests * Cleanup test errors
davidpcls
left a comment
There was a problem hiding this comment.
Comments to assist others reviewing the code
| # Reject any of those old sessions and force reauthentication. | ||
| return None | ||
|
|
||
| session = db.query(Session).filter(Session.uuid == uuid_module.UUID(hex=session_id)).first() |
There was a problem hiding this comment.
This does not request eager loading like Tiled. The data structure of Tiled does eager loading of the pending session automatically and since we're still on the sync version of the database this is fine.
| On successful lookup the matching ``Identity.latest_login`` is updated to | ||
| now. | ||
| """ | ||
| identity = db.query(Identity).filter(Identity.id == id).filter(Identity.provider == identity_provider).first() |
There was a problem hiding this comment.
This does not request eager loading like Tiled. The data structure of Tiled does eager loading of the pending session automatically and since we're still on the sync version of the database this is fine.
| """ | ||
| hashed_device_code = hashlib.sha256(device_code).digest() | ||
| pending_session = ( | ||
| db.query(PendingSession).filter(PendingSession.hashed_device_code == hashed_device_code).first() |
There was a problem hiding this comment.
This does not request eager loading like Tiled. The data structure of Tiled does eager loading of the pending session automatically and since we're still on the sync version of the database this is fine.
| @@ -0,0 +1,37 @@ | |||
| from abc import ABC | |||
There was a problem hiding this comment.
File copied from Tiled
| @@ -0,0 +1,80 @@ | |||
| #!/usr/bin/env bash | |||
There was a problem hiding this comment.
I'll fully admit, i did not write this myself. AI did it. It works reliably for me and brings test times down for local development to around 10 minutes.
There was a problem hiding this comment.
I'll fully admit, i did not write this myself. AI did it. It works reliably for me and brings test times down for local development to around 10 minutes.
There was a problem hiding this comment.
This looks like a good idea to run local tests in multiple independent processes. It should also be possible to use pytest-split instead of shards to balance load between processes (as in https://github.com/bluesky/bluesky-queueserver/blob/0a084cfb722308f384847c24e6a43a8f572f302e/.github/workflows/testing.yml#L50). This way the tests are split on equal groups based on execution time. It seems to work well in the bluesky-queueserver repository for running CI tests. But this is for future discussion.
|
After these changes, still seems to be working for me. Neato. |




Description
These changes bring in the authenticators from Tiled 0.2.12 into bluesky-httpserver along with the required changes to the app and supporting files to support them. This allows us to use the same authentication setup as from Tiled deployments (at least at the current moment). The main other change is to remove the "mode" flag and instead use the class type to determine if it is an internal or external authenticator.
Part of this work is also the updating of the bluesky-queueserver-api, for which I have created a PR. Both are required in order for the OIDC workflows to work.
Like Tiled, the authentication flow forces a login prompt on each attempt to login. I have tried to minimize any differences between similar functions/files in Tiled and bluesky-httpserver.
Motivation and Context
This solves the problem of having different authentication schemes to maintain between Tiled and HTTP server, which came from the same code around 3 years ago. Tiled has been updated but HTTP server was not. This addresses that.
Summary of Changes for Release Notes
Updated authenticators based off Tiled v0.2.12.
Made minimal changes to app.py and authentication.py to support the changes. Some new endpoints needed to be created to do this, but fairly minimal. The majority of the work was getting the unit test runners stable.
Added local parallel unit test runners to aid development.
Updated the github runners to be more stable and reliable by making the port used dynamic.
Fixed
Added
Changed
Removed
How Has This Been Tested?
Testing was done against MS Entra. I tested that these workflows work for both internal (localhost) and external servers. Using this the login workflow changes to just simply being:
While this was tested with MS Entra, the design is based on Tiled's OIDC and testing used the OIDC flow with MS Entra, so it should theoretically work with all other authenticators using OIDC.
TODO: