coulomb-social/docs/release/personal-login-500-2026-10-06.md
tegwick 1292539069
Some checks failed
Use-case regression gate / usecase-regressions (push) Has been cancelled
docs: record reset receipt and personal login 500 deployment correction
Assistant: codex
Assistant-Model: gpt-6.1-sol
Assistant-Session: 01a10e2d-a5ef-7e92-b03d-b1694e915680
2026-10-06 20:52:27 +02:00

2.7 KiB

Personal login HTTP500 correction — 2026-10-06

The user confirmed receipt of a fresh platform password-reset email, then reported Server Error500 after authentication. End-to-end personal app login is not yet accepted. CSOC-WP-0008-T11 and CSOC-WP-0010-T03 remain wait.

Diagnosis

The app's identity-linking client requests /api/v1/me on USER_ENGINE_BASE_URL. The deployed historical address https://users.92-205-62-239.nip.io returns302 to https://users.coulomb.social/api/v1/me. httpx does not follow redirects by default; raise_for_status rejects302. establish_session calls that client before recording last_login_at or establishing the Django session; its failure is uncaught by the callback. This explains the reported symptom; raw traceback was not available from the production app's recent stdout log.

Credential-safe in-pod probes send an empty claims envelope (no subject/issuer) using the configured service credential internally: old address302; canonical address400 missing claims. This establishes the canonical address accepts the service credential without creating an identity, sending email, exporting credentials or impersonating the user. It does not prove full signed-in use.

Correction deployed

Owning railiance-apps main commit3f8276e changes USER_ENGINE_BASE_URL in chart defaults and production overrides to https://users.coulomb.social. Helm lint passed. Node-side server dry-run and parsed manifest comparison passed, proving only that environment value changes. Image, Secret references, issuer/client/ tenant, PVC, access rules and every other manifest field remain unchanged.

This completes the personal app deployment already approved by the founder (“Go on and deploy the interruption is not a problem.”), decision 8bbce691-a787-4140-9cd4-df06e8635a27. Authority: ADMINISTER @ realm:kubernetes/railiance01, activation=APPROVED. Node IP verified92.205.62.239. Node-side Helm --reuse-values --atomic --wait --timeout5m succeeded: release coulomb-social revision11 deployed, chart0.2.0, same image digest sha256:bf975939e22b9b0a5d9ac05584ae6265ba201169df6dc48e506ec684b30a9c58. One of one replicas available. Public /healthz and landing200, login302. Post-deploy in-pod empty-claims probe confirms canonical base URL,400 response and no redirect using the existing service credential (probe exit0). Rollback target: Helm revision10, which restores the historical URL.

Acceptance

The user was asked to start a fresh sign-in at https://app.coulomb.social/auth/login/ and report workspace access or error. Only that successful interactive flow closes personal-login acceptance. Inbox receipt is accepted separately. No private protocol was imported and Bubble remains on the apex.