HTTPError: 502 Server Error triggered by repair_faulty_edx_user, non-ascii username

[sentry-io]

Sentry Issue: XPRO-5ZG

HTTPError: 502 Server Error: Bad Gateway for url: https://courses.xpro.mit.edu/auth/complete/mitxpro-oauth2/?redirect_state=5O8jLzykCZksmXK8Zc63qbgU69yc40sO&code=PVvo0FieLNiF19hQlWusWrh50P9miV&state=5O8jLzykCZksmXK8Zc63qbgU69yc40sO
  File "courseware/api.py", line 317, in repair_faulty_courseware_users
    created_user, created_auth_token = repair_faulty_edx_user(user)
  File "courseware/api.py", line 297, in repair_faulty_edx_user
    created_auth_token = create_edx_auth_token(user) is not None
  File "contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File "courseware/api.py", line 163, in create_edx_auth_token
    resp.raise_for_status()
  File "requests/models.py", line 960, in raise_for_status
    raise HTTPError(http_error_msg, response=self)

Failed to repair faulty user Response - code: 502, url: https://courses.xpro.mit.edu/auth/complete/mitxpro-oauth2/?redirect_state=... content: (HTML body ignored)

[mbertrand]

Response content from edx server is HTML with text:

Uh oh, we are having some server issues..
Server error
We are aware of the error and are working to find a resolution.

from courseware.api.create_edx_auth_token:

        url = edx_url(OPENEDX_SOCIAL_LOGIN_XPRO_PATH)
        resp = req_session.get(url)
        resp.raise_for_status()

[mbertrand]

Appears to be related to non-ascii "full name".

To recreate:

• Register as a new user. When filling out the "Full Name" field, use the same value as in the linked sentry error. Use regular ascii characters for first and last name fields.
• Purchase a course/program
• Click the edx link to the course. Instead of loading, it shows a server error page.

[mbertrand]

When creating the user in local xpro with devstack:

web_1     | [2023-02-14 15:44:32] ERROR 15 [root] user.py:282 - [c15bbfb46fcc] - Error creating courseware user records on User create
web_1     | Traceback (most recent call last):
web_1     |   File "/src/./authentication/pipeline/user.py", line 280, in create_courseware_user
web_1     |     courseware_api.create_user(user)
web_1     |   File "/src/./courseware/api.py", line 70, in create_user
web_1     |     create_edx_user(user)
web_1     |   File "/src/./courseware/api.py", line 114, in create_edx_user
web_1     |     raise CoursewareUserCreateError(
web_1     | courseware.exceptions.CoursewareUserCreateError: Error creating Open edX user. Response - code: 400, content: {
web_1     |   "username": [
web_1     |     {
web_1     |       "user_message": "Usernames can only contain letters (A-Z, a-z), numerals (0-9), underscores (_), and hyphens (-)."
web_1     |     }
web_1     |   ],
web_1     |   "error_code": "validation-error"
web_1     | }

False alarm, does not happen on RC or prod. Must be a local devstack setting somewhere.

[mbertrand]

Error occurs with usernames having non-ascii characters, but there does not seem to be a standard length beyond which the error is raised.

A sample name in Chinese characters only works when the number of unicode characters is shortened to a length of 10:
奕泽-奕泽-奕泽-奕

Adding one more Chinese character causes the error.

But a sample name in Arabic characters works when the number of unicode characters is twice that length.