9.6 KiB
Multi-Address Single Deployment Data Implementation Plan
For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (
- [ ]) syntax for tracking.
Goal: Ensure external and local/LAN access addresses are documented and verified as multiple entrypoints to the same deployment stack and database, not two separately running data sets.
Architecture: The application keeps one active Compose stack for formal business data. External and local addresses both route to that stack's Nginx, which forwards /api/ to the same NestJS server and PostgreSQL volume. docker-compose.local.yml remains an independent local/LAN demo stack and must not be described as a synced copy of production data.
Tech Stack: Docker Compose, Nginx reverse proxy template, Next.js same-origin API config, Node deployment verifier, Markdown deployment docs.
File Structure
- Modify
scripts/verify-production-deploy.mjs: add checks that deployment docs and env examples describe multi-address single-stack behavior. - Modify
.env.production.example: add comments thatSERVER_NAMEis the primary public host and local/LAN access must route to the same Nginx stack. - Modify
.env.local-server.example: add comments that this is an independent local/LAN demo stack, not a synced view of production data. - Modify
docs/deployment.md: explain the recommended multi-address single deployment and warn against running prod/local stacks for the same business. - Modify
docs/architecture.md: document that one formal business stack may have multiple access addresses. - Modify
docs/decisions.md: add a decision that address-level consistency is solved by one active deployment stack, not data sync. - Modify
docs/workflow.md: update deployment flow and troubleshooting checks for address routing.
Task 1: Add Deployment Verifier Expectations
Files:
-
Modify:
scripts/verify-production-deploy.mjs -
Step 1: Add failing verifier snippets
In scripts/verify-production-deploy.mjs, update existing checks so the deployment verifier requires the new wording.
Add these snippets to the .env.production.example check:
'Primary public host',
'same Compose Nginx',
Add these snippets to the .env.local-server.example check:
'independent local/LAN demo stack',
'does not sync with production data',
Add these snippets to the docs/deployment.md check:
'同一部署多入口访问',
'外网地址和本地地址',
'同一套 Compose',
'不要同时启动',
Add these snippets to the docker-compose.prod.yml check:
'SERVER_NAME',
'NEXT_PUBLIC_API_URL',
Keep all existing checks that still apply.
- Step 2: Run verifier to confirm it fails
Run:
pnpm deploy:verify
Expected: command exits non-zero and prints missing snippets for .env.production.example, .env.local-server.example, and docs/deployment.md.
- Step 3: Commit failing verifier
Run:
git add scripts/verify-production-deploy.mjs
git commit -m "test(deploy): 覆盖多入口单部署校验"
Task 2: Clarify Environment Examples
Files:
-
Modify:
.env.production.example -
Modify:
.env.local-server.example -
Step 1: Update production env comments
In .env.production.example, replace:
# Public entrypoint
SERVER_NAME=pm.example.com
with:
# Public and local entrypoints
# Primary public host. Local/LAN addresses should route to the same Compose Nginx stack,
# not to a separate docker-compose.local.yml stack, when they need the same business data.
SERVER_NAME=pm.example.com
Keep NEXT_PUBLIC_API_URL=/api/v1.
- Step 2: Update local-server env comments
In .env.local-server.example, replace:
# Copy to .env.local-server for an all-in-one local/LAN deployment.
with:
# Copy to .env.local-server for an independent local/LAN demo stack.
# This stack uses local_* volumes and does not sync with production data.
# If external and local addresses must show the same business data, route both addresses to the same production Compose stack instead.
- Step 3: Run verifier to confirm docs still fail
Run:
pnpm deploy:verify
Expected: command still exits non-zero because docs/deployment.md has not yet been updated.
- Step 4: Commit env comments
Run:
git add .env.production.example .env.local-server.example
git commit -m "docs(deploy): 标注本地栈不与生产数据同步"
Task 3: Update Deployment And Architecture Docs
Files:
-
Modify:
docs/deployment.md -
Modify:
docs/architecture.md -
Modify:
docs/decisions.md -
Modify:
docs/workflow.md -
Step 1: Add deployment section for multi-address access
In docs/deployment.md, add this section before “本地服务器部署”:
## 同一部署多入口访问
如果只是希望外网地址和本地地址看到同一份业务数据,不要启动两套部署栈。正确方式是让外网域名、公网 IP、内网 IP 或本机地址都进入同一套 Compose 的 Nginx。
```text
外网地址和本地地址
-> 同一套 Compose Nginx
-> 同一个 web/server
-> 同一个 PostgreSQL volume
```
生产同域部署保持:
```env
NEXT_PUBLIC_API_URL=/api/v1
```
这样浏览器从哪个地址打开页面,就请求该地址下的 `/api/v1`。只要这些地址最终进入同一个 Nginx/server,数据就是同一份。
不要同时启动 `docker-compose.prod.yml` 和 `docker-compose.local.yml` 来承载同一套正式业务。它们使用不同 volume,会看到两套数据。
Use the four-backtick outer fence shown above if copying this plan into another Markdown document, because the section contains nested fences.
- Step 2: Clarify local-server deployment is independent
In docs/deployment.md, under “本地服务器部署”, add this paragraph near the top:
本地服务器部署是一套独立 local/LAN 演示栈,使用 `local_*` volumes。它适合办公室内网试用或离线演示,不会与云服务器生产数据自动保持一致。如果目标是外网地址和本地地址看到同一份正式业务数据,请使用上面的“同一部署多入口访问”,而不是同时运行两套栈。
- Step 3: Update architecture production boundary
In docs/architecture.md, under the production deployment layer, add:
同一套正式业务部署可以有多个访问入口,例如外网域名、公网 IP、内网 IP 或本机地址。多入口必须进入同一个 Nginx/server/PostgreSQL 栈,才能看到同一份数据。`docker-compose.local.yml` 是独立本地/LAN 演示栈,不能作为生产栈的数据同步副本。
- Step 4: Add decision record
In docs/decisions.md, append a new decision:
## 40. 外网与本地访问地址通过同一部署栈保持数据一致
**问题**:用户希望外网地址和本地/内网地址访问系统时看到同一份业务数据。如果两个地址分别打到 `docker-compose.prod.yml` 和 `docker-compose.local.yml`,就会使用不同 PostgreSQL volume,数据天然分叉。
**决策**:外网地址和本地地址应作为同一套正式业务部署的多个入口,全部路由到同一个 Nginx/server/PostgreSQL 栈。`docker-compose.local.yml` 只用于独立本地/LAN 演示,不承诺与生产数据一致。
**理由**:这是访问入口问题,不是数据同步问题。保持单部署栈可以复用现有 `/api/v1` 同域配置和 `app_data` 乐观锁,避免引入双向同步、冲突合并和数据库复制复杂度。
- Step 5: Update workflow deployment flow
In docs/workflow.md, in the production deployment flow, add:
外网地址和本地地址要看到同一份数据时,必须让两个地址都进入同一套 Compose/Nginx。不要同时用 `docker-compose.prod.yml` 和 `docker-compose.local.yml` 承载同一套业务;本地栈使用 `local_*` volumes,只适合独立演示。
Also add this troubleshooting item:
如果外网地址和本地地址数据不一致,先检查两个地址是否打到了同一个 `/api/v1/config/ai` 后端和同一个 Compose project;不要先做数据库同步。
- Step 6: Run verifier and type checks
Run:
pnpm deploy:verify
pnpm --filter web type-check
pnpm --filter server type-check
Expected:
-
pnpm deploy:verifyprintsProduction deployment artifacts verified. -
Both type-check commands exit 0.
-
Step 7: Commit docs
Run:
git add docs/deployment.md docs/architecture.md docs/decisions.md docs/workflow.md
git commit -m "docs(deploy): 说明多入口访问同一部署数据"
Task 4: Final Verification And Reporting
Files:
-
No file changes expected.
-
Step 1: Run final verification
Run:
pnpm deploy:verify
pnpm --filter web type-check
pnpm --filter server type-check
git status --short
Expected:
-
pnpm deploy:verifyexits 0. -
pnpm --filter web type-checkexits 0. -
pnpm --filter server type-checkexits 0. -
git status --shortshows only intentional plan/spec files if they were not committed, otherwise no output. -
Step 2: Report operational guidance
In the final response, include:
外网地址和本地地址要看同一份数据时,只运行一套正式业务 Compose;让所有地址都反代到同一个 Nginx/server/PostgreSQL。
Also include:
docker-compose.local.yml 是独立演示栈;如果同时和生产栈运行,会看到两套数据。