psa-systems/mokosh-server
Watch
9
0
Fork
You've already forked mokosh-server
 1
Code Issues Pull requests Projects Releases 10 Packages Wiki Activity Actions

ci(e2e): route API calls through the SPA-host proxy while PMS-618 is open #423

Closed
YousifShkara wants to merge 2 commits from chore/PMS-618-e2e-api-base-workaround into main
pull from: chore/PMS-618-e2e-api-base-workaround
merge into: psa-systems:main
Conversation 0 Commits 2 Files changed -
YousifShkara commented 2026-07-02 11:25:51 +02:00
Owner
Copy link

The mokosh-server PR CI runs e2e/scripts/health-check.mjs as a preflight (PMS-141) and deriveApiBase prepends api. to the SPA host, landing on https://api.msp.a8n.systems/api/v1/health. That subdomain is currently 404-ing on staging: the Traefik router labels for the api host are missing, so every mokosh-server PR fails at the preflight step (PR #421 and #422 both showed Staging /api/v1/health returned HTTP 404). Same failure mode as the CORS-404 storm on the user's /tickets screenshot: the SPA-derived api host does not resolve to mokosh-server.

Root-cause fix (Traefik + DNS) lives outside this repo and is tracked in PMS-618. Land the short-term unblock here so the queue moves: set E2E_API_BASE_URL to the same value as E2E_BASE_URL (the SPA host), which serves /api/v1/* correctly today via the SPA-host proxy that already exists. deriveApiBase is skipped in favour of the explicit override, so the preflight and every /api/v1/* test call go to a working host. env.opBaseURL is unaffected because it has its own explicit source (E2E_OP_BASE_URL / OIDC_ISSUER_*).

Remove this override in the same PR that closes PMS-618: the workaround should not silently outlive the routing fix. The comment on the env line points at the ticket to make that removal easy to find.

#PMS-618

The mokosh-server PR CI runs `e2e/scripts/health-check.mjs` as a preflight (PMS-141) and `deriveApiBase` prepends `api.` to the SPA host, landing on `https://api.msp.a8n.systems/api/v1/health`. That subdomain is currently 404-ing on staging: the Traefik router labels for the api host are missing, so every mokosh-server PR fails at the preflight step (PR #421 and #422 both showed `Staging /api/v1/health returned HTTP 404`). Same failure mode as the CORS-404 storm on the user's /tickets screenshot: the SPA-derived api host does not resolve to mokosh-server. Root-cause fix (Traefik + DNS) lives outside this repo and is tracked in PMS-618. Land the short-term unblock here so the queue moves: set `E2E_API_BASE_URL` to the same value as `E2E_BASE_URL` (the SPA host), which serves `/api/v1/*` correctly today via the SPA-host proxy that already exists. `deriveApiBase` is skipped in favour of the explicit override, so the preflight and every /api/v1/* test call go to a working host. `env.opBaseURL` is unaffected because it has its own explicit source (`E2E_OP_BASE_URL` / `OIDC_ISSUER_*`). Remove this override in the same PR that closes PMS-618: the workaround should not silently outlive the routing fix. The comment on the env line points at the ticket to make that removal easy to find. #PMS-618
ci(e2e): route API calls through the SPA-host proxy while PMS-618 is open
Some checks failed
E2E / Playwright against staging (pull_request) Failing after 1m28s
Details
Check / fmt + clippy + build + tests (pull_request) Successful in 3m38s
Details
Integration / integration tests (pull_request) Has been cancelled
Details
65ee68fb4f 
The mokosh-server PR CI runs `e2e/scripts/health-check.mjs` as a preflight (PMS-141) and `deriveApiBase` prepends `api.` to the SPA host, landing on `https://api.msp.a8n.systems/api/v1/health`. That subdomain is currently 404-ing on staging: the Traefik router labels for the api host are missing, so every mokosh-server PR fails at the preflight step (PR #421 and #422 both showed `Staging /api/v1/health returned HTTP 404`). Same failure mode as the CORS-404 storm on the user's /tickets screenshot: the SPA-derived api host does not resolve to mokosh-server.

Root-cause fix (Traefik + DNS) lives outside this repo and is tracked in PMS-618. Land the short-term unblock here so the queue moves: set `E2E_API_BASE_URL` to the same value as `E2E_BASE_URL` (the SPA host), which serves `/api/v1/*` correctly today via the SPA-host proxy that already exists. `deriveApiBase` is skipped in favour of the explicit override, so the preflight and every /api/v1/* test call go to a working host. `env.opBaseURL` is unaffected because it has its own explicit source (`E2E_OP_BASE_URL` / `OIDC_ISSUER_*`).

Remove this override in the same PR that closes PMS-618: the workaround should not silently outlive the routing fix. The comment on the env line points at the ticket to make that removal easy to find.

#PMS-618
Revert "ci(e2e): route API calls through the SPA-host proxy while PMS-618 is open"
Some checks failed
E2E / Playwright against staging (pull_request) Failing after 6s
Details
Check / fmt + clippy + build + tests (pull_request) Successful in 1m26s
Details
Integration / integration tests (pull_request) Successful in 5m11s
Details
Create release / Gate (release-branch merges only) (pull_request) Successful in 1s
Details
Create release / Create release from merged PR (pull_request) Has been skipped
Details
ec8298f88b 
The workaround was based on a wrong reading of the staging state. I confirmed with `curl` that `https://msp.a8n.systems/api/v1/health` returns 200 and thought that was the SPA-host proxy answering. It's not - the 200 was the mokosh SPA's index.html fallback, served for every unknown path (any-URL 200 with HTML body is exactly how a static SPA host looks). The `/api` prefix at that host is not a proxy to mokosh-server at all.

PR #423 CI (the one this workaround shipped on) confirmed it: `E2E_API_BASE_URL=https://msp.a8n.systems` cleared the preflight (that same 200 fooled `health-check.mjs`, which only inspects `res.ok`), then every real test failed. POST returned 405 (the static SPA host refuses POST) and GET returned HTML the teardown could not parse ("Unexpected token '<'...").

Net effect of the workaround: preflight green, entire test surface red. Strictly worse than the original preflight-red state, because a green preflight masks how broken staging actually is.

Reverting the env override so the CI is honest again: preflight fails at `Staging /api/v1/health returned HTTP 404`, which correctly points at PMS-618 (staging Traefik / DNS for `api.msp.<tld>` is missing). Fixing PMS-618 in the docker repo is the only real unblock; there is no other host that answers `/api/v1/*` from the outside on staging right now.

#PMS-618
YousifShkara closed this pull request 2026-07-03 12:42:01 +02:00
YousifShkara deleted branch chore/PMS-618-e2e-api-base-workaround 2026-07-03 12:42:05 +02:00
Some checks failed
E2E / Playwright against staging (pull_request) Failing after 6s
Required
Details
Check / fmt + clippy + build + tests (pull_request) Successful in 1m26s
Required
Details
Integration / integration tests (pull_request) Successful in 5m11s
Details
Create release / Gate (release-branch merges only) (pull_request) Successful in 1s
Details
Create release / Create release from merged PR (pull_request) Has been skipped
Details

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
YousifShkara
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
psa-systems/mokosh-server!423
No description provided.