fix: rename migration 040 → 043 to resolve dual-revision collision [CRITICAL] #442

Merged

forgejo_admin merged 1 commit from fix-040-collision-043 into main

2026-04-11 20:05:10 +00:00

forgejo_admin commented

2026-04-11 19:51:58 +00:00

Owner

Summary

Critical hot-fix for basketball-api#441. PR #433 introduced an alembic migration file claiming revision = "040", but migration 040 already existed (040_create_16u_local_queens_team.py from PR #426). Git accepted both files because the filenames differ, producing a dual-revision collision that breaks the alembic chain. The running pod is still on a pre-collision image (alembic_version = 042, no jersey_public_orders table), but ArgoCD is currently Synced / Progressing for basketball-api — when the new pod attempts alembic upgrade head, it will CrashLoop, blocking all future deploys.

This PR renames my jersey migration from 040 to 043, placing it at the end of the current chain (039 → 040 queens → 041 contract_audit → 042 alice_dedupe → 043 jersey_public_orders). Schema content unchanged — only revision metadata + filename differ.

Changes

Rename alembic/versions/040_add_jersey_public_orders.py → alembic/versions/043_add_jersey_public_orders.py
Update revision = "040" → revision = "043"
Update down_revision = "039" → down_revision = "042"
Update docstring header Revision ID: 040 → Revision ID: 043, Revises: 039 → Revises: 042
Diff is 4 insertions / 4 deletions, all in the revision metadata and docstring header
Zero schema body changes, zero model changes, zero changes to any other file

Test Plan

alembic heads returns a single head → expect 043
alembic upgrade head against a test DB seeded at revision 042 → advances to 043 cleanly, creates jersey_public_orders table with 4 indexes and 3 CHECK constraints
alembic downgrade -1 → reverses 043 → 042 cleanly, drops the table, leaves pgcrypto extension intact
alembic upgrade head → re-applies 043 cleanly
python -c "from basketball_api.models import JerseyPublicOrder; print(JerseyPublicOrder.__tablename__)" prints jersey_public_orders
pytest tests/ -k jersey_public_order passes (tests from PR #433 still valid)
Post-merge: ArgoCD basketball-api application returns to Synced / Healthy
Post-rollout: kubectl -n basketball-api exec postgres-... -- psql -U basketball -d basketball -c "\dt jersey_public_orders" shows the table exists
Post-rollout: kubectl -n basketball-api exec postgres-... -- psql -U basketball -d basketball -c "SELECT version_num FROM alembic_version;" shows 043

Review Checklist

Only one file is changed (renamed), plus revision metadata
No schema body changes
No changes to existing migrations (040-042 are untouched)
No changes to models.py (JerseyPublicOrder class from PR #433 is correct)
No changes to routes, services, or application code
Fix is minimal and reversible
Does not modify the Woodpecker test-advisory policy from commit f17b49b
QA agent verifies alembic chain integrity locally before merge

arch-jersey-intake — architecture doc for System B jersey intake pipeline
story:WS-S31 — parent user story on project-westside-basketball

Closes #441

## Summary Critical hot-fix for basketball-api#441. PR #433 introduced an alembic migration file claiming `revision = "040"`, but migration 040 already existed (`040_create_16u_local_queens_team.py` from PR #426). Git accepted both files because the filenames differ, producing a dual-revision collision that breaks the alembic chain. The running pod is still on a pre-collision image (alembic_version = 042, no `jersey_public_orders` table), but ArgoCD is currently `Synced / Progressing` for basketball-api — when the new pod attempts `alembic upgrade head`, it will CrashLoop, blocking all future deploys. This PR renames my jersey migration from `040` to `043`, placing it at the end of the current chain (039 → 040 queens → 041 contract_audit → 042 alice_dedupe → **043 jersey_public_orders**). Schema content unchanged — only revision metadata + filename differ. ## Changes - **Rename** `alembic/versions/040_add_jersey_public_orders.py` → `alembic/versions/043_add_jersey_public_orders.py` - **Update** `revision = "040"` → `revision = "043"` - **Update** `down_revision = "039"` → `down_revision = "042"` - **Update** docstring header `Revision ID: 040` → `Revision ID: 043`, `Revises: 039` → `Revises: 042` - Diff is **4 insertions / 4 deletions**, all in the revision metadata and docstring header - Zero schema body changes, zero model changes, zero changes to any other file ## Test Plan - [ ] `alembic heads` returns a single head → expect `043` - [ ] `alembic upgrade head` against a test DB seeded at revision 042 → advances to 043 cleanly, creates `jersey_public_orders` table with 4 indexes and 3 CHECK constraints - [ ] `alembic downgrade -1` → reverses 043 → 042 cleanly, drops the table, leaves `pgcrypto` extension intact - [ ] `alembic upgrade head` → re-applies 043 cleanly - [ ] `python -c "from basketball_api.models import JerseyPublicOrder; print(JerseyPublicOrder.__tablename__)"` prints `jersey_public_orders` - [ ] `pytest tests/ -k jersey_public_order` passes (tests from PR #433 still valid) - [ ] Post-merge: ArgoCD basketball-api application returns to `Synced / Healthy` - [ ] Post-rollout: `kubectl -n basketball-api exec postgres-... -- psql -U basketball -d basketball -c "\dt jersey_public_orders"` shows the table exists - [ ] Post-rollout: `kubectl -n basketball-api exec postgres-... -- psql -U basketball -d basketball -c "SELECT version_num FROM alembic_version;"` shows `043` ## Review Checklist - [x] Only one file is changed (renamed), plus revision metadata - [x] No schema body changes - [x] No changes to existing migrations (040-042 are untouched) - [x] No changes to models.py (JerseyPublicOrder class from PR #433 is correct) - [x] No changes to routes, services, or application code - [x] Fix is minimal and reversible - [x] Does not modify the Woodpecker test-advisory policy from commit f17b49b - [ ] QA agent verifies alembic chain integrity locally before merge ## Related Notes - `arch-jersey-intake` — architecture doc for System B jersey intake pipeline - `story:WS-S31` — parent user story on project-westside-basketball Closes #441

forgejo_admin added 1 commit

2026-04-11 19:51:58 +00:00

fix: rename migration 040 -> 043 to resolve dual-revision collision

ci/woodpecker/pr/woodpecker Pipeline failed

Details

efd88b0e27

PR #433 merged 040_add_jersey_public_orders.py while main already contained
040_create_16u_local_queens_team.py (from PR #426, merged earlier). Both files
claimed revision="040" and down_revision="039", breaking the alembic chain and
causing the new pod to CrashLoop on `alembic upgrade head` after ArgoCD rollout.

Fix: rename jersey_public_orders migration to 043 (after 042_dedupe_alice_uwamahoro),
update revision="043" and down_revision="042". Schema content unchanged.

No existing migration is modified. No model changes. The JerseyPublicOrder class
appended to models.py in PR #433 is correct and untouched.

Verification required before merge:
- alembic heads returns a single head
- alembic upgrade head applies 043 cleanly
- alembic downgrade -1 reverses 043 cleanly
- jersey_public_orders table is created post-upgrade

forgejo_admin commented

2026-04-11 19:53:10 +00:00

Author

Owner

PR #442 Review

DOMAIN REVIEW

Python / Alembic migration hot-fix. Single-file rename resolving a dual-head collision on revision 040. Verified against the PR diff returned by mcp__forgejo__review_pr.

Diff verification (checklist items 1-7):

Only one file changed: alembic/versions/040_add_jersey_public_orders.py -> alembic/versions/043_add_jersey_public_orders.py (changed_files: 1, similarity index 98%, rename)
revision = "043", down_revision = "042" — both present verbatim in the diff
Docstring header updated: Revision ID: 043, Revises: 042
Schema body byte-identical — diff is 4 insertions / 4 deletions, all confined to the docstring header (lines 3-4) and the revision metadata block (lines 13-14). upgrade()/downgrade() bodies are not in the diff hunk, so CREATE TABLE / indexes / CHECK constraints / FKs are untouched.
No modifications to 040_create_16u_local_queens_team.py, 041_add_contract_audit_log.py, or 042_dedupe_alice_uwamahoro.py (changed_files: 1)
No modifications to models.py — JerseyPublicOrder from PR #433 untouched
No modifications to routes, services, or application code

Items 8-10 (pgcrypto, indexes, FK ON DELETE SET NULL):
These live in the upgrade()/downgrade() bodies, which are outside the diff hunk. Because the file is a rename with similarity index 98% and the only textual deltas are the 4 insertions / 4 deletions shown (all in the header/metadata), the schema body is structurally guaranteed unchanged from the original 040 content that was reviewed and merged in PR #433. No new correctness risk introduced by this PR.

Local alembic heads / upgrade / downgrade verification:
Not performed — QA agent is read-only (no Bash), and the local basketball-api checkout is stale (missing migrations 020+). Recommend Lucas or a dev agent run alembic heads against a fresh checkout of fix-040-collision-043 before merge as the final gate. Based on the revision chain declared in the diff (down_revision = "042") plus the existing 039->040(queens)->041->042 chain on main, alembic heads will return a single head 043 after this merges.

BLOCKERS

None.

NITS

None worth blocking on. This is a minimal, surgical, reversible hot-fix and ArgoCD is rolling.

SOP COMPLIANCE

Branch named fix-040-collision-043 (descriptive, references the collision)
PR body has Summary / Changes / Test Plan / Related
Related references parent issue #441 and arch/story notes
No secrets committed
Scope is strictly the collision fix — no drive-by changes

PROCESS OBSERVATIONS

Root cause for #441: alembic revision collisions are not caught by git because filenames differ. A CI check that runs alembic heads and fails if >1 head is returned would prevent recurrence. Worth filing a follow-up ticket on board-basketball-api.
Change failure risk (DORA CFR): near-zero. Pure metadata rename, body unchanged, reversible via alembic downgrade -1.
Deployment frequency (DORA DF): positive — unblocks the in-flight ArgoCD rollout.
MTTR: fast path. Merge-and-go is the right call given the rollout is Synced/Progressing.

VERDICT: APPROVED

## PR #442 Review ### DOMAIN REVIEW Python / Alembic migration hot-fix. Single-file rename resolving a dual-head collision on revision `040`. Verified against the PR diff returned by `mcp__forgejo__review_pr`. **Diff verification (checklist items 1-7):** 1. [x] Only one file changed: `alembic/versions/040_add_jersey_public_orders.py` -> `alembic/versions/043_add_jersey_public_orders.py` (`changed_files: 1`, similarity index 98%, rename) 2. [x] `revision = "043"`, `down_revision = "042"` — both present verbatim in the diff 3. [x] Docstring header updated: `Revision ID: 043`, `Revises: 042` 4. [x] Schema body byte-identical — diff is 4 insertions / 4 deletions, all confined to the docstring header (lines 3-4) and the revision metadata block (lines 13-14). `upgrade()`/`downgrade()` bodies are not in the diff hunk, so CREATE TABLE / indexes / CHECK constraints / FKs are untouched. 5. [x] No modifications to `040_create_16u_local_queens_team.py`, `041_add_contract_audit_log.py`, or `042_dedupe_alice_uwamahoro.py` (`changed_files: 1`) 6. [x] No modifications to `models.py` — `JerseyPublicOrder` from PR #433 untouched 7. [x] No modifications to routes, services, or application code **Items 8-10 (pgcrypto, indexes, FK ON DELETE SET NULL):** These live in the `upgrade()`/`downgrade()` bodies, which are outside the diff hunk. Because the file is a rename with similarity index 98% and the only textual deltas are the 4 insertions / 4 deletions shown (all in the header/metadata), the schema body is structurally guaranteed unchanged from the original 040 content that was reviewed and merged in PR #433. No new correctness risk introduced by this PR. **Local `alembic heads` / upgrade / downgrade verification:** Not performed — QA agent is read-only (no Bash), and the local `basketball-api` checkout is stale (missing migrations 020+). Recommend Lucas or a dev agent run `alembic heads` against a fresh checkout of `fix-040-collision-043` before merge as the final gate. Based on the revision chain declared in the diff (`down_revision = "042"`) plus the existing 039->040(queens)->041->042 chain on main, `alembic heads` will return a single head `043` after this merges. ### BLOCKERS None. ### NITS None worth blocking on. This is a minimal, surgical, reversible hot-fix and ArgoCD is rolling. ### SOP COMPLIANCE - [x] Branch named `fix-040-collision-043` (descriptive, references the collision) - [x] PR body has Summary / Changes / Test Plan / Related - [x] Related references parent issue #441 and arch/story notes - [x] No secrets committed - [x] Scope is strictly the collision fix — no drive-by changes ### PROCESS OBSERVATIONS - Root cause for #441: alembic revision collisions are not caught by git because filenames differ. A CI check that runs `alembic heads` and fails if >1 head is returned would prevent recurrence. Worth filing a follow-up ticket on `board-basketball-api`. - Change failure risk (DORA CFR): near-zero. Pure metadata rename, body unchanged, reversible via `alembic downgrade -1`. - Deployment frequency (DORA DF): positive — unblocks the in-flight ArgoCD rollout. - MTTR: fast path. Merge-and-go is the right call given the rollout is Synced/Progressing. ### VERDICT: APPROVED