Launch blocker: validate the release candidate on real AWS S3 with end-to-end Git hosting

未關閉
  • 哪裡不對？
#222 2 則留言 0 個 reaction 已指派 1 人 在 GitHub 檢視

維護者通常 1 天內回覆

@ThomasK33 已經在處理了。

開始於 2026年9月11日。

評估

這個 Issue 還沒有評估資料。

描述

Launch gate

Before the public open-source launch and Show HN announcement, record a reproducible real-AWS validation of the release candidate. SeaweedFS and Floci provide valuable regression coverage, but they do not establish that AWS authentication, permissions, storage behavior, and the complete Git hosting flow work together in the intended deployment.

Existing evidence and remaining gap

This is not the first AWS test. The recorded S3 campaign from August 27, 2026 passed the 22-surface qualification suite and exercised real Git pushes and recovery against S3. It was a single-day campaign from compute outside AWS and predates subsequent changes.

The testing reference documents emulator gaps: the pinned Floci lane does not enforce S3 IAM policy decisions or fully validate header-auth signatures/session tokens. Its passing Git/LFS suite therefore cannot prove those AWS properties.

The existing S3-only backend promotion tracker owns experimental-to-supported criteria, including pending contention reruns and multi-day measurements. This issue owns the product-wide public-launch AWS validation; it does not promote S3-only storage or replace that tracker.

Acceptance criteria

  • Define an isolated AWS sandbox, dedicated scratch resources, region, least-privilege test role, spending limit, and reliable teardown before provisioning. Use short-lived credentials and retain no secrets in evidence.
  • Pin the exact release candidate commit and binary/container artifact. Provide reproducible setup and test commands, preferably automated, with logs and a clear pass/fail report.
  • Run from EC2 in the same AWS region as the test bucket. Cover the standard Postgres + S3 topology, including two server replicas; separately exercise the advertised experimental single-node S3-only mode and keep its status explicit.
  • Verify repository creation and API reads/writes, Git clone/fetch/push, branches/tags, authorization denials, Git LFS uploads/downloads, and agreement between API and Git views. Verify that writes through one standard-mode replica can be read through the other.
  • Test AWS-specific behavior that emulators cannot establish: valid and invalid temporary credentials, session-token handling/refresh, denied IAM permissions, expected bucket owner, valid/tampered presigned URLs, and the conditional-write semantics used by S3-only storage.
  • Exercise concurrent writes, cold-cache restart, interrupted push/recovery, and the existing S3-only qualification and lease/restart drills on real S3. Verify acknowledged repository data remains readable and report any failed or skipped case explicitly. Link remaining S3-only promotion criteria to the existing tracker.
  • If Kubernetes/EKS is part of the launch support claim, add a focused deployment smoke covering workload identity, ingress/TLS, pod replacement, and rollout behavior. Otherwise record EKS as a follow-up rather than a prerequisite for proving S3 compatibility.
  • Publish the evidence and supported deployment limits, resolve release-blocking failures, and verify resource teardown. Close this launch blocker only when the release candidate has the required evidence.

Scope

Keep emulator CI coverage. Real AWS testing supplements it. This issue tracks readiness; it does not authorize provisioning resources, incurring cloud spend, changing repository visibility, or publishing a release.

主要語言
Go
星號
0
分支
0
平均合併
6 小時 16 分鐘
30 天內合併 PR
243

環境準備

  • 提供 Dockerfile 或 Docker Compose 檔案
  • 沒有 Pull Request 範本
  • 閱讀貢獻指南

從這裡開始

  1. 先讀完整個 Issue，再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫，在一個分支上完成修改。
  4. 送出 Pull Request，並在描述裡引用這個 Issue 編號。

coder/openstorage 的其他 Issue

查看 coder/openstorage 的全部 Issue

相似的 Issue

更多 Go Issue

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。