cal/claude-memory
1
0
Fork 0
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

store: Fix: Prevent duplicate POSTs when audit update fails after successful delivery in outbound-event-handler

Browse Source
This commit is contained in:
Cal Corum 2026-02-25 11:47:36 -06:00
parent 671d7069d3
commit 9c4889d00f
1 changed files with 46 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
graph/fixes/fix-prevent-duplicate-posts-when-audit-update-fails-after-su-fccd42.md Normal file
View File

@ -0,0 +1,46 @@
---
id: fccd4277-a253-4f7f-8065-7a9cffd64354
type: fix
title: "Fix: Prevent duplicate POSTs when audit update fails after successful delivery in outbound-event-handler"
tags: [esb-monorepo, tac, outbound-event-handler, error-handling, fix, cloud-functions, idempotency, duplicate-prevention]
importance: 0.7
confidence: 0.8
created: "2026-02-25T17:47:36.634745+00:00"
updated: "2026-02-25T17:47:36.634745+00:00"
---
# Fix: Duplicate POST Prevention on Audit Failure
## Project
esb-monorepo — TAC `outbound-event-handler` (GCF)
## Problem
Bitbucket Rovo Dev flagged a code design issue: if `update_processed_events()` fails at Step 5 (after all POSTs to `outbound-object-router` succeeded), the function returned HTTP 500. Cloud Scheduler treats any 5xx as a failure and retries, causing duplicate POSTs to the downstream router for events that were already delivered.
## Root Cause
The final audit DB update call (`update_processed_events`) was not wrapped in error handling. A transient DB failure at that step caused the function to surface a 500, even though all delivery work was complete.
## Fix
Wrapped the final `update_processed_events` call in a `try/except` block:
- Logs the audit error at ERROR level
- Still returns **200** because all deliveries succeeded — the audit failure is non-fatal
- Cloud Scheduler will NOT retry, preventing duplicate POSTs
## Important Distinction
The partial-progress flush in the POST failure path (~line 165) is intentionally **NOT** wrapped. In that case we DO want 500 returned so Cloud Scheduler retries the failed POST delivery.
```python
# Step 5 — audit update: non-fatal, deliveries already complete
try:
update_processed_events(processed_ids)
except Exception as e:
logger.error("Audit update failed after successful delivery: %s", e)
return make_response("OK", 200)
```
## Test Added
`test_audit_update_failure_after_all_posts_succeed_returns_200`
## Commit
`320b871` on `dev-object-handler` branch