8000 Enforce single-writer deployment settings by noobydp · Pull Request #240 · tracefinity/tracefinity · GitHub
[go: up one dir, main page]

Skip to content

Enforce single-writer deployment settings - #240

Draft
noobydp wants to merge 1 commit into
tracefinity:mainfrom
noobydp:fix/single-writer-deployment
Draft

noobydp wants to merge 1 commit into
tracefinity:mainfrom
noobydp:fix/single-writer-deployment

Conversation

@noobydp
@noobydp noobydp commented Oct 25, 2026
Copy link
Copy Markdown
Contributor

Multiple backend processes sharing a storage volume can overwrite acknowledged saves because the JSON stores and generation locks are process-local. Pin the container to one Uvicorn worker even when a platform sets WEB_CONCURRENCY, reject Helm replica counts above one, and document the single-writer requirement.

The chart retains Recreate so a one-replica upgrade does not briefly run two writers. Source launches and separate containers sharing a volume remain an operator constraint; this change does not implement multi-process storage.

Closes #155 (the issue's minimum single-worker deployment scope).

Validation:

  • Complete backend suite: 858 passed.
  • Complete frontend suite: 275 passed.
  • make lint: backend lint, frontend lint and TypeScript passed; 9 existing frontend warnings.
  • Six deployment regressions exercise the actual Uvicorn CLI/config precedence under WEB_CONCURRENCY=4 and render the chart with default, 0, 1, 2 and 4 replicas. Helm runs in backend CI.

Validation ran on macOS with Python 3.12, the repository-declared pnpm 10.29.2, and Helm 4.3.1. A full Docker image runtime smoke test was not run locally.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Investigate running the backend as multiple processes or replicas

1 participant

0