isoastra/pipeline-service
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix: pipeline robustness hardening #1

Closed
ronit wants to merge 0 commits from fix/pipeline-robustness-hardening into master
pull from: fix/pipeline-robustness-hardening
merge into: isoastra:master
Conversation 0 Commits - Files changed -
ronit commented 2026-03-25 02:01:38 +00:00
Owner
Copy link

Summary

  • Atomic webhook dedup — replace TOCTOU DashMap check-then-insert with entry() API, preventing duplicate pipeline runs from concurrent webhooks
  • Dedup eviction — background task clears stale entries every 30s, preventing unbounded memory growth
  • NATS flush — completion events now call flush() after publish() to guarantee delivery
  • RPC timeout — switch to request_rpc_with_timeout(300s) instead of unbounded request_rpc
  • Reviewer — fetch default branch from Forgejo API instead of hardcoding main/master
  • ORG_ACCOUNT_ID — deduplicate inline UUID functions into shared constant
  • Silent failure logging — 5 previously silent failure paths now log warnings (missing PR, malformed repo name, missing event header)
  • Pre-initial migration — 202603229999_create_pipeline_schema.sql ensures CREATE SCHEMA IF NOT EXISTS pipeline runs before the initial migration on fresh databases

Test plan

  • cargo check — compiles with pre-existing warnings only
  • Deployed to prod — pod healthy, webhook processing confirmed via logs
  • New migration applied successfully alongside existing migrations

🤖 Generated with Claude Code

## Summary - **Atomic webhook dedup** — replace TOCTOU DashMap check-then-insert with `entry()` API, preventing duplicate pipeline runs from concurrent webhooks - **Dedup eviction** — background task clears stale entries every 30s, preventing unbounded memory growth - **NATS flush** — completion events now call `flush()` after `publish()` to guarantee delivery - **RPC timeout** — switch to `request_rpc_with_timeout(300s)` instead of unbounded `request_rpc` - **Reviewer** — fetch default branch from Forgejo API instead of hardcoding main/master - **ORG_ACCOUNT_ID** — deduplicate inline UUID functions into shared constant - **Silent failure logging** — 5 previously silent failure paths now log warnings (missing PR, malformed repo name, missing event header) - **Pre-initial migration** — `202603229999_create_pipeline_schema.sql` ensures `CREATE SCHEMA IF NOT EXISTS pipeline` runs before the initial migration on fresh databases ## Test plan - [x] `cargo check` — compiles with pre-existing warnings only - [x] Deployed to prod — pod healthy, webhook processing confirmed via logs - [x] New migration applied successfully alongside existing migrations 🤖 Generated with [Claude Code](https://claude.com/claude-code)
fix: harden pipeline-service against silent failures and race conditions
All checks were successful
pipeline-service No pipeline action (save)
6f23a97c1f 
- webhook dedup: replace TOCTOU DashMap check-then-insert with atomic entry() API
- webhook dedup: add background eviction task (30s interval, 60s TTL) to prevent unbounded growth
- notify: add nats.client.flush() after publish to ensure completion events reach NATS
- deploy RPC: switch to request_rpc_with_timeout (300s) instead of unbounded request_rpc
- deploy: use shared ORG_ACCOUNT_ID constant instead of per-file inline functions
- reviewer: fetch default branch from Forgejo API instead of hardcoding main/master
- executor: log warnings for missing PR (result was silently dropped) and malformed repo names
- webhook: log warnings for missing/non-ASCII x-forgejo-event header and malformed repo names
- migration: add compensating CREATE SCHEMA IF NOT EXISTS for bare sqlx migrate safety

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
fix: replace compensating migration with pre-initial schema creation
All checks were successful
pipeline-service No pipeline action (save)
e81796a370 
The previous compensating migration (202603240001) ran after the initial
migration and could not protect bare `sqlx migrate run` on a fresh database.
Replace with 202603229999 which runs before the initial migration, creating
the pipeline schema before SET search_path TO pipeline executes.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
fix: restore applied migration file — sqlx requires it in the resolved set
All checks were successful
pipeline-service No pipeline action (save)
1bd1800025 
sqlx validates that all previously-applied migrations still exist as files.
Deleting 202603240001 caused CrashLoopBackOff. Restore the file while
keeping 202603229999 as the pre-initial schema creation for fresh DBs.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
fix: restore exact original migration content to match sqlx checksum
All checks were successful
pipeline-service No pipeline action (save)
da8bd60a7f 
sqlx checksums the full file content including comments. The restored
file had different comments, causing "migration has been modified" error.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ronit closed this pull request 2026-03-25 04:57:31 +00:00
All checks were successful
pipeline-service No pipeline action (save)

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
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
isoastra/pipeline-service!1
No description provided.