docs(ops): record 103 goosed native migration and add migrate tooling
Memind CI / Test, build, and release guards (pull_request) Has been cancelled
Memind CI / Test, build, and release guards (pull_request) Has been cancelled
Document the completed Docker-to-native goosed pool cutover on 103, add the production migrate script with backup/rollback paths, and extend local native pool soak/metrics helpers used as migration gates. Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
@@ -1,6 +1,6 @@
|
||||
# Local and 103 runtime topology
|
||||
|
||||
> Last confirmed: 2026-07-26 20:18 CST.
|
||||
> Last confirmed: 2026-07-30 20:20 CST (goosed native migration).
|
||||
>
|
||||
> This is the current topology source of truth for local Memind and 103 production. Prefer this document over older migration notes. Older architecture documents may contain historical paths from before the MindSpace split.
|
||||
|
||||
@@ -75,48 +75,62 @@ or whose `gitSha` does not match the runtime artifact manifest.
|
||||
|
||||
## goosed
|
||||
|
||||
103 goosed runs in Colima/Docker, not as native launchd processes.
|
||||
103 goosed runs as **native launchd** processes (2026-07-30 migrated). Docker/Colima pool is stopped; `goosed-prod-1` is kept as emergency standby.
|
||||
|
||||
> 迁移记录与回滚:[goosed-native-103-migration-plan.md](./goosed-native-103-migration-plan.md)(**已执行 2026-07-30**)。脚本:`scripts/goosed-native-103-migrate.sh`。
|
||||
|
||||
| Item | Value |
|
||||
|------|-------|
|
||||
| Docker context | `colima` |
|
||||
| Compose directory | `/Users/john/Project/goosed-prod` |
|
||||
| Compose env | `/Users/john/Project/goosed-prod/.env` |
|
||||
| Image pattern | `tkmind/goosed:prod-${GOOSED_TAG}` |
|
||||
| Session store | PostgreSQL `memind_sessions` through `GOOSE_SESSION_DB_URL` |
|
||||
| Session PostgreSQL | **Host** `postgresql@17` on `127.0.0.1:5432` (`/opt/homebrew/var/postgresql@17`) |
|
||||
| Runtime type | Native launchd (`GOOSED_RUNTIME=native`) |
|
||||
| Root | `/Users/john/Project/tkmind_go-native` |
|
||||
| Binary | `tkmind_go-native/goosed` → `releases/goosed-*` |
|
||||
| Launchd labels | `cn.tkmind.goosed-native-18006` … `18014` |
|
||||
| Run script | `run-goosed-native.sh <port>` + `.env.<port>` |
|
||||
| Session store | PostgreSQL `memind_sessions` via `GOOSE_SESSION_DB_URL` → `127.0.0.1:5432` |
|
||||
| FD monitor | `cn.tkmind.goosed-monitor` → `scripts/monitor-goosed-fds.mjs` |
|
||||
| Docker standby | `goosed-prod-1` stopped, `restart=no`(紧急回滚 18006) |
|
||||
| DeepSeek no-think | `cn.tkmind.memind-deepseek-no-think` → `:18036`(native 必需,`MEMIND_DEEPSEEK_PROXY_ENTRYPOINT=1`) |
|
||||
|
||||
**Critical:** Colima goosed reaches the session DB via `host.docker.internal:5432`. If host PostgreSQL is stopped, H5 `/agent/start` hangs ~60s then fails with `pool timed out`, and the UI stays on「正在创建新对话…」. Recovery:
|
||||
**Critical:** Native goosed uses host PostgreSQL on `127.0.0.1:5432`. If PostgreSQL stops, H5 `/agent/start` hangs ~60s. Recovery:
|
||||
|
||||
```bash
|
||||
ssh john@58.38.22.103 'bash /Users/john/Project/goosed-prod/scripts/ensure-goose-session-postgres.sh'
|
||||
ssh john@58.38.22.103 'bash /Users/john/Project/Memind/scripts/ensure-goose-session-postgres.sh'
|
||||
ssh john@58.38.22.103 '/opt/homebrew/opt/postgresql@17/bin/pg_isready -h 127.0.0.1 -p 5432'
|
||||
```
|
||||
|
||||
`cn.tkmind.goosed-monitor`(103)运行 `goosed-prod/scripts/monitor-goosed-containers.mjs`,本地 native goosed 监控用 `scripts/monitor-goosed-fds.mjs`;两者都会在巡检时调用 `ensure-goose-session-postgres.sh`。
|
||||
Port layout (native listens directly on host port):
|
||||
|
||||
Container layout (host port → container `18006`):
|
||||
|
||||
| Container | Host port | Container port |
|
||||
|-----------|-----------|----------------|
|
||||
| `goosed-prod-1` | `18006` | `18006` |
|
||||
| `goosed-prod-2` | `18007` | `18006` |
|
||||
| `goosed-prod-3` | `18008` | `18006` |
|
||||
| `goosed-prod-4` | `18009` | `18006` |
|
||||
| `goosed-prod-5` | `18010` | `18006` |
|
||||
| `goosed-prod-6` | `18011` | `18006` |
|
||||
| `goosed-prod-7` | `18012` | `18006` |
|
||||
| `goosed-prod-8` | `18013` | `18006` |
|
||||
| `goosed-prod-9` | `18014` | `18006` |
|
||||
| Port | Launchd label |
|
||||
|------|---------------|
|
||||
| `18006` | `cn.tkmind.goosed-native-18006` |
|
||||
| `18007` | `cn.tkmind.goosed-native-18007` |
|
||||
| … | … |
|
||||
| `18014` | `cn.tkmind.goosed-native-18014` |
|
||||
|
||||
All nine host ports must return `ok` from `/status`, and must appear in Portal `.env` `TKMIND_API_TARGETS`.
|
||||
|
||||
### Legacy Docker pool (standby only)
|
||||
|
||||
| Item | Value |
|
||||
|------|-------|
|
||||
| Compose directory | `/Users/john/Project/goosed-prod` |
|
||||
| Image pattern | `tkmind/goosed:prod-${GOOSED_TAG}` |
|
||||
| Containers | `goosed-prod-1..9` — **stopped**, `docker update --restart=no` |
|
||||
|
||||
Do **not** leave `cn.tkmind.goosed-monitor` on `monitor-goosed-containers.mjs` with `GOOSED_RESTART_UNHEALTHY=1`; it will restart Docker and fight native ports.
|
||||
|
||||
```bash
|
||||
# Emergency: single Docker instance on 18006 only
|
||||
launchctl bootout gui/$(id -u)/cn.tkmind.goosed-native-18006
|
||||
docker update --restart=unless-stopped goosed-prod-1 && docker start goosed-prod-1
|
||||
```
|
||||
|
||||
## PostgreSQL rule
|
||||
|
||||
Both local native goosed and 103 Colima goosed persist sessions in PostgreSQL:
|
||||
Both local native goosed and 103 native goosed persist sessions in PostgreSQL:
|
||||
|
||||
- local native goosed: PostgreSQL database `goose_sessions_dev`
|
||||
- 103 Colima goosed: PostgreSQL database `memind_sessions`
|
||||
- 103 native goosed: PostgreSQL database `memind_sessions`
|
||||
|
||||
Do not assume a local SQLite session store when debugging current Memind/goosed behavior.
|
||||
|
||||
|
||||
@@ -0,0 +1,365 @@
|
||||
# 103 生产 goosed 迁移计划:Docker/Colima → Native 多实例
|
||||
|
||||
> **状态:** 已执行(2026-07-30 维护窗)
|
||||
> **适用范围:** 103(`58.38.22.103` / Mac Studio)生产环境
|
||||
> **目标运行时:** 去掉 goosed 对 Docker/Colima 的依赖,改为 native launchd 多实例(`18006`–`18014`)
|
||||
> **前提:** 允许维护窗口停机;**不要求**进行中 SSE/工具任务不断线
|
||||
> **关联拓扑:** [103-runtime-topology.md](./103-runtime-topology.md)
|
||||
> **发布闸门:** [production-release-guardian.md](./production-release-guardian.md)、[发包必看.md](./发包必看.md)
|
||||
|
||||
---
|
||||
|
||||
## 1. 目标与边界
|
||||
|
||||
### 1.1 要达成什么
|
||||
|
||||
| 目标 | 说明 |
|
||||
|------|------|
|
||||
| 去掉 goosed Docker/Colima 池 | `goosed-prod-1..9` + canary 容器退役 |
|
||||
| 改为 native launchd 池 | 参考 103 已有 `tkmind_go-native` / 本机 `install-local-goosed-pool.sh` 模式,扩到 9 实例 |
|
||||
| 用户数据零丢失 | RDS、PostgreSQL、MindSpace 磁盘原样保留 |
|
||||
| 功能等价恢复 | 登录、聊天、续聊、页面生成、微信、Plaza/adm 等与迁移前一致 |
|
||||
| 消除 bind mount inode 漂移 | Portal 发版后不再依赖 `docker compose force-recreate` remount |
|
||||
|
||||
### 1.2 「无缝」的定义(本计划)
|
||||
|
||||
| 维度 | 本计划承诺 | 不承诺 |
|
||||
|------|------------|--------|
|
||||
| 账号、余额、MindSpace 页面、历史落库 | 零丢失 | — |
|
||||
| 系统功能 | 维护窗结束后等价恢复 | — |
|
||||
| 进行中的 SSE / 工具执行 | — | 不断线;用户需刷新或重开对话 |
|
||||
| 零停机 | — | 必须安排维护窗口 |
|
||||
|
||||
架构约束([memind-2-streaming-agent-runtime-plan.md](./architecture/memind-2-streaming-agent-runtime-plan.md)):
|
||||
|
||||
- 已存在 session **保持 worker affinity,不做中途迁移**
|
||||
- **不做无损中途迁移**,除非 goosed 支持完整 session restore
|
||||
|
||||
### 1.3 不在本次范围
|
||||
|
||||
- 103 迁到新物理机
|
||||
- 实例数从 9 缩到 2(首次迁移禁止同时缩容)
|
||||
- 修改 RDS / 用户 schema
|
||||
- 合并尚未走完发布闸门的 Portal 大功能(除非维护窗内单独批准发版)
|
||||
|
||||
---
|
||||
|
||||
## 2. 生产数据与依赖地图
|
||||
|
||||
```text
|
||||
105 nginx (m.tkmind.cn)
|
||||
│
|
||||
▼
|
||||
103 Portal :8081 cn.tkmind.memind-portal
|
||||
│
|
||||
├── 阿里云 RDS MySQL `goose` 用户、计费、h5_user_sessions.goosed_target、agent_run…
|
||||
├── MindSpace Service :8082 /Users/john/MindSpace(不动)
|
||||
├── memind_adm :8085 (不动)
|
||||
├── Plaza :3001 (不动)
|
||||
│
|
||||
├── goosed 18006–18014 【本次替换】Docker → native
|
||||
│ └── PG5432 memind_sessions(共享,不迁移数据)
|
||||
│
|
||||
├── PG5433 mindspace_userdata_prod
|
||||
├── Redis :6379 运行时路由/指标(可清空重建)
|
||||
└── 磁盘(Portal persist)
|
||||
.env
|
||||
/Users/john/MindSpace/data/mindspace
|
||||
/Users/john/Project/Memind/users
|
||||
/Users/john/Project/Memind/data
|
||||
public/plaza-covers, logs, .tailscale
|
||||
```
|
||||
|
||||
### 2.1 持久化清单(必须备份)
|
||||
|
||||
与 `scripts/release-portal-runtime-prod.sh` manifest 一致:
|
||||
|
||||
```text
|
||||
.env, MindSpace, data, users, .tailscale, public/plaza-covers, logs
|
||||
```
|
||||
|
||||
**额外必须备份(不在 persist tar 内):**
|
||||
|
||||
| 资产 | 备份方式 |
|
||||
|------|----------|
|
||||
| 阿里云 RDS `goose` | 控制台快照 + 可选逻辑导出 |
|
||||
| PostgreSQL `memind_sessions` | `pg_dump` |
|
||||
| PostgreSQL `mindspace_userdata_prod` | `pg_dump` |
|
||||
| `goosed-prod` compose + `.env` + 镜像 tag | 目录 tar + `GOOSED_TAG` 记录 |
|
||||
| `tkmind_go-native`(若已存在) | 目录拷贝 |
|
||||
| Portal `.release-manifest.txt` | 记录 `release_id` / `git_head` |
|
||||
| 105 nginx `m.tkmind.cn` 等配置 | 发布脚本同款备份 |
|
||||
|
||||
### 2.2 Session 与流量(迁移后行为不变)
|
||||
|
||||
- **新对话:** Portal `pickTarget()` 在 `TKMIND_API_TARGETS` 上轮询 / Redis 负载评分
|
||||
- **老对话:** MySQL `h5_user_sessions.goosed_target` + Redis `session:{id}:target` 粘住原 `https://127.0.0.1:1800N`
|
||||
- **goosed 会话体:** PG `memind_sessions` 多实例共享;native 直连 `127.0.0.1:5432`(替代 Docker `host.docker.internal`)
|
||||
|
||||
**首次迁移硬规则:native 仍监听 `18006`–`18014` 全端口**,与现网 `goosed_target` URL 兼容。
|
||||
|
||||
---
|
||||
|
||||
## 3. 迁移前准入条件(Gate)
|
||||
|
||||
以下 **全部满足** 才允许申请 103 维护窗:
|
||||
|
||||
### 3.1 本机验证(不动 103)
|
||||
|
||||
| # | 项 | 命令 / 标准 |
|
||||
|---|-----|-------------|
|
||||
| 1 | native 双实例 soak | `node scripts/soak-local-goosed-pool.mjs --minutes 30 --interval 60` 通过 |
|
||||
| 2 | 池健康检查 | `node scripts/check-local-goosed-pool.mjs` |
|
||||
| 3 | Portal 多 target | `.env` 中 `TKMIND_API_TARGETS=18006,18007` + 续聊 / 新建 session 冒烟 |
|
||||
| 4 | native 指标 | `GOOSED_RUNTIME=native node scripts/runtime-worker-metrics.mjs status --dry-run` |
|
||||
| 5 | 安装脚本 | `bash scripts/install-local-goosed-pool.sh` 可重复执行 |
|
||||
| 6 | 发布脚本 native 分支 | `GOOSED_RUNTIME=native` 时 `release-portal-runtime-prod.sh` 不依赖 Docker remount(合并 main 后) |
|
||||
| 7 | 回滚演练 | 本机或 103 只读:Docker compose 停/起 + `/status` 全绿 |
|
||||
|
||||
### 3.2 仓库与发布
|
||||
|
||||
| # | 项 |
|
||||
|---|-----|
|
||||
| 1 | 相关脚本 / 文档已合并 `main`,CI 通过 |
|
||||
| 2 | `bash scripts/check-release-ready.sh` 通过 |
|
||||
| 3 | 生产发布守门员 Core + Impact Gate report 已准备(与维护窗发版 commit 绑定) |
|
||||
| 4 | 用户 **明确批准** 103 维护窗与迁移执行(口头「继续」不构成批准) |
|
||||
|
||||
### 3.3 103 只读基线(迁移前 24h 内采集)
|
||||
|
||||
```bash
|
||||
# 只读,禁止改 103
|
||||
ssh john@58.38.22.103 'cat /Users/john/Project/Memind/.release-manifest.txt'
|
||||
ssh john@58.38.22.103 'grep ^TKMIND_API_TARGETS= /Users/john/Project/Memind/.env'
|
||||
ssh john@58.38.22.103 'grep ^GOOSED_TAG= /Users/john/Project/goosed-prod/.env'
|
||||
ssh john@58.38.22.103 'for p in $(seq 18006 18014); do curl -kfsS --connect-timeout 2 https://127.0.0.1:$p/status; echo; done'
|
||||
ssh john@58.38.22.103 'curl -fsS http://127.0.0.1:8082/mindspace/v1/contract | head -c 400'
|
||||
```
|
||||
|
||||
保存到本机 `test/_103_baselines/pre-native-migration-YYYYMMDD/`。
|
||||
|
||||
---
|
||||
|
||||
## 4. 维护窗口执行计划
|
||||
|
||||
**建议窗口:** ≥ 30 分钟(9 实例安装 + provider 同步 + 回归)
|
||||
**建议时段:** 低峰 + 可联系 105 改 nginx 的人员在场
|
||||
|
||||
### Phase 0 — 备份(T0 前,可先执行)
|
||||
|
||||
- [ ] RDS 快照
|
||||
- [ ] `pg_dump memind_sessions`、`mindspace_userdata_prod`
|
||||
- [ ] 103 上执行 persist 备份(或等价 `memind-persisted-*-before.tar.gz`)
|
||||
- [ ] 全目录 / compose / 105 nginx 备份
|
||||
- [ ] 记录:`release_id`、`GOOSED_TAG`、九个 Docker 容器 ID、`TKMIND_API_TARGETS`
|
||||
|
||||
### Phase 1 — 切流量(停机开始)
|
||||
|
||||
- [ ] 105:维护页或 upstream 指向维护(禁止新用户进入)
|
||||
- [ ] 103:可选公告 / 微信模板(「系统维护约 N 分钟」)
|
||||
- [ ] **Soft drain(推荐,最多等 10 分钟):**
|
||||
|
||||
```bash
|
||||
# 103 上,对每个 worker 设 drain,等待 active_streams 归零
|
||||
# 见 memind-2-streaming-agent-runtime-plan.md P5 drain 说明
|
||||
docker exec memind-runtime-redis redis-cli SET memind:runtime:worker:goosed-N:drain 1
|
||||
curl -sk https://m.tkmind.cn/api/runtime/status # 观察 activeStreams
|
||||
```
|
||||
|
||||
- [ ] 超时后 **硬停**:`cn.tkmind.memind-portal`(或仅阻止新 session,视 drain 结果)
|
||||
|
||||
### Phase 2 — 替换 goosed 运行时
|
||||
|
||||
- [ ] 停止 Docker goosed 池:
|
||||
|
||||
```bash
|
||||
cd /Users/john/Project/goosed-prod
|
||||
docker compose -f docker-compose.prod.yml down
|
||||
docker rm -f goosed-prod-canary 2>/dev/null || true
|
||||
```
|
||||
|
||||
- [ ] 部署 native 9 实例(目录模式参考 `tkmind_go-native`,端口 `18006`–`18014`):
|
||||
- 同一 `GOOSE_SESSION_DB_URL` → `memind_sessions`
|
||||
- `GOOSE_TLS=true`
|
||||
- `GOOSED_MCP_CONTAINERIZED=0`
|
||||
- `GOOSED_MCP_*` 使用宿主机路径(与 Portal `.env` 一致)
|
||||
- `MINDSPACE_STORAGE_ROOT=/Users/john/MindSpace/data/mindspace`
|
||||
- `PORTAL_RUNTIME_DIR=/Users/john/Project/Memind`
|
||||
|
||||
- [ ] 更新 Portal `.env`(若需要):
|
||||
- `TKMIND_API_TARGETS=https://127.0.0.1:18006,...,https://127.0.0.1:18014`(**含 18006**)
|
||||
- `GOOSED_RUNTIME=native`
|
||||
- DeepSeek no-think:`127.0.0.1:18036`(去掉 `host.docker.internal` 依赖)
|
||||
|
||||
- [ ] **MindSpace :8082、RDS、磁盘:不移动、不删**
|
||||
|
||||
- [ ] Provider 同步到 **全部 9 个 target**
|
||||
|
||||
- [ ] 验收九个 `/status`:
|
||||
|
||||
```bash
|
||||
for p in $(seq 18006 18014); do
|
||||
echo -n "$p: "
|
||||
curl -kfsS --connect-timeout 3 "https://127.0.0.1:$p/status" || echo FAIL
|
||||
done
|
||||
```
|
||||
|
||||
### Phase 3 — Portal 与附属服务
|
||||
|
||||
**若维护窗内无 Portal 代码变更:**
|
||||
|
||||
- [ ] 重启 `cn.tkmind.memind-portal`(读新 `.env`)
|
||||
- [ ] **跳过** Docker remount 步骤
|
||||
|
||||
**若维护窗内包含 Portal runtime 发版:**
|
||||
|
||||
- [ ] 按 [发包必看.md](./发包必看.md) 走 canary 或晋升流程
|
||||
- [ ] 发版脚本使用 `GOOSED_RUNTIME=native` 检查链
|
||||
- [ ] MindSpace contract `gitSha` 与 manifest 对齐
|
||||
|
||||
**始终确认仍在运行:**
|
||||
|
||||
- [ ] `cn.tkmind.mindspace-service` :8082
|
||||
- [ ] memind_adm :8085、Plaza :3001、imgproxy、SearXNG、agent-run-worker、DeepSeek no-think :18036
|
||||
|
||||
### Phase 4 — 功能回归(103 localhost + 105 外网)
|
||||
|
||||
| 优先级 | 场景 | 验证方式 |
|
||||
|--------|------|----------|
|
||||
| P0 | Portal `/api/status` | `curl http://127.0.0.1:8081/api/status` |
|
||||
| P0 | 登录 / 鉴权 | `/auth/login`、`/auth/status` |
|
||||
| P0 | 新建对话 | H5 开聊,检查 `h5_user_sessions.goosed_target` 分布 |
|
||||
| P0 | **旧对话续聊** | 维护前存在的 session,迁移后 `/reply` 成功 |
|
||||
| P0 | 页面列表 / 公开页 | MindSpace remote + 已有 HTML 可访问 |
|
||||
| P1 | page.generate / edit_file | workspace 落盘 + `public/*.html` |
|
||||
| P1 | 微信真实回调 | 非仅 health |
|
||||
| P1 | MindSearch / 图片 imgproxy | 生产 URL |
|
||||
| P1 | agent-run worker | 队列消费 |
|
||||
| P2 | Plaza、adm 后台 | 各端口 health |
|
||||
|
||||
生产 Gate:维护窗发版 commit 须绑定 **Core + Impact Gate report**(见 production-release-guardian)。
|
||||
|
||||
### Phase 5 — 恢复流量
|
||||
|
||||
- [ ] 105 nginx 恢复 `58.38.22.103:8081`(或既定 upstream)
|
||||
- [ ] 外网抽样:`https://m.tkmind.cn/api/status`
|
||||
- [ ] 维护结束公告
|
||||
|
||||
### Phase 6 — 迁移后观察(24–72h)
|
||||
|
||||
- [ ] `GOOSED_RUNTIME=native node scripts/runtime-worker-metrics.mjs sample`
|
||||
- [ ] native FD 监控(`monitor-goosed-fds.mjs`,覆盖 `18006`–`18014`)
|
||||
- [ ] Redis worker 指标、SLO report 无异常尖刺
|
||||
- [ ] 用户续聊 / 页面生成无集中投诉
|
||||
|
||||
---
|
||||
|
||||
## 5. 回滚计划
|
||||
|
||||
**触发条件:** P0 回归失败、九个 target 无法全绿、旧 session 批量无法续聊、FD 泄漏导致整机不稳定
|
||||
|
||||
**目标时间:** 维护窗内 < 15 分钟完成回滚
|
||||
|
||||
| 步骤 | 操作 |
|
||||
|------|------|
|
||||
| 1 | 105 维护页保持 |
|
||||
| 2 | 停 native launchd 池(`18006`–`18014`) |
|
||||
| 3 | `cd goosed-prod && docker compose -f docker-compose.prod.yml up -d`(**原 GOOSED_TAG**) |
|
||||
| 4 | `.env` 恢复备份:`TKMIND_API_TARGETS`、Docker MCP 路径、`GOOSED_RUNTIME=docker` |
|
||||
| 5 | 若 Portal 已发新版:回滚至备份 `release_id` runtime |
|
||||
| 6 | 九容器 healthy + provider 同步 + `/status` 全绿 |
|
||||
| 7 | 105 恢复流量 |
|
||||
|
||||
**回滚不需要:** 恢复 RDS / PG(迁移中若未改库)。
|
||||
|
||||
---
|
||||
|
||||
## 6. 配置对照表(Docker → Native)
|
||||
|
||||
| 配置项 | Docker 生产(现) | Native 生产(目标) |
|
||||
|--------|-------------------|---------------------|
|
||||
| goosed 进程 | `goosed-prod-1..9` 容器 | `com.tkmind.goosed-native-1800N` launchd |
|
||||
| Session PG | `host.docker.internal:5432` | `127.0.0.1:5432` |
|
||||
| MCP | 容器内 `/usr/local/bin/node` 等 | 宿主机 `GOOSED_MCP_*` |
|
||||
| `GOOSED_MCP_CONTAINERIZED` | 隐式容器 | `0` 或不设 |
|
||||
| Portal remount | 发版后 `compose force-recreate` | **不需要** |
|
||||
| 监控 | `monitor-goosed-containers.mjs` | `monitor-goosed-fds.mjs`(扩展全端口) |
|
||||
| 指标 | `runtime-worker-metrics` docker | `GOOSED_RUNTIME=native` |
|
||||
| Canary goosed | `goosed-prod-canary:18015` | native `18015`(Phase 2 后单独迁移) |
|
||||
|
||||
---
|
||||
|
||||
## 7. 风险与缓解
|
||||
|
||||
| 风险 | 严重度 | 缓解 |
|
||||
|------|--------|------|
|
||||
| 进行中对话全部中断 | 高 | 维护公告;soft drain;恢复后提示刷新 |
|
||||
| 旧 session PG restore 失败 | 高 | 迁移前用 PG 副本在 staging 验续聊 |
|
||||
| provider 未覆盖全 target | 高 | 自动化检查 + 发版脚本 gate |
|
||||
| MCP 路径 / sandbox 失败 | 高 | `page.generate` smoke + sandbox-fs |
|
||||
| `goosed_target` 指向已下线端口 | 中 | 首次迁移保持 9 端口不变 |
|
||||
| native FD 泄漏 ×9 | 中 | 72h 监控 + kickstart 阈值 |
|
||||
| MindSpace / Portal 版本漂移 | 中 | contract gitSha 对账 |
|
||||
| 105 切流后 CDN/缓存 | 低 | 外网真实 URL 验证 |
|
||||
|
||||
---
|
||||
|
||||
## 8. 时间线模板
|
||||
|
||||
| 时点 | 动作 | 环境 |
|
||||
|------|------|------|
|
||||
| T-14d | 本机 native 双实例 → 脚本 native 化完成 | 本机 |
|
||||
| T-7d | 本机 native 9 实例 soak(若与生产同构) | 本机 |
|
||||
| T-3d | 103 只读基线 + 备份演练 + Docker 回滚演练 | 103 只读/演练 |
|
||||
| T-1d | 维护公告定稿;105 维护页脚本就绪 | 105 |
|
||||
| T0 | Phase 0–5 执行 | 103 + 105 |
|
||||
| T+24h | FD / 续聊 / 微信观察 | 103 |
|
||||
| T+7d | 评审:是否退役 Colima、删 Docker 镜像、更新 topology 文档 | 文档 |
|
||||
| T+30d | 可选:评估实例缩容(**单独变更**,不在本计划内) | — |
|
||||
|
||||
---
|
||||
|
||||
## 9. 迁移后文档更新
|
||||
|
||||
迁移成功并稳定 7 天后:
|
||||
|
||||
- [ ] 更新 [103-runtime-topology.md](./103-runtime-topology.md)(goosed 段改为 native 9 实例)
|
||||
- [ ] 更新 [发包必看.md](./发包必看.md) §5 goosed 镜像 → native 二进制发布
|
||||
- [ ] 在 [docs/branch-disposition.md](./branch-disposition.md) 登记相关功能分支(若适用)
|
||||
- [ ] 归档 Docker compose 回滚包位置与 `GOOSED_TAG`
|
||||
|
||||
---
|
||||
|
||||
## 10. 本仓库工具索引
|
||||
|
||||
| 用途 | 路径 |
|
||||
|------|------|
|
||||
| 本机安装 native 池 | `bash scripts/install-local-goosed-pool.sh` |
|
||||
| 池健康检查 | `node scripts/check-local-goosed-pool.mjs` |
|
||||
| 长跑 soak | `node scripts/soak-local-goosed-pool.mjs` |
|
||||
| native worker 指标 | `GOOSED_RUNTIME=native node scripts/runtime-worker-metrics.mjs` |
|
||||
| FD 监控 | `scripts/monitor-goosed-fds.mjs` + `scripts/install-goosed-monitor.sh` |
|
||||
| Portal 生产发布 | `scripts/release-portal-runtime-prod.sh` |
|
||||
| 发布就绪 | `bash scripts/check-release-ready.sh` |
|
||||
|
||||
---
|
||||
|
||||
## 11. 批准记录(执行时填写)
|
||||
|
||||
| 字段 | 值 |
|
||||
|------|-----|
|
||||
| 维护窗口 | YYYY-MM-DD HH:MM – HH:MM (UTC+8) |
|
||||
| 执行人 | |
|
||||
| Portal `git_head` | |
|
||||
| Native goosed 二进制 / release | |
|
||||
| 回滚 `GOOSED_TAG` | |
|
||||
| RDS 快照 ID | |
|
||||
| persist_backup 路径 | |
|
||||
| Gate report ID | |
|
||||
| 结果 | 成功 / 回滚 |
|
||||
| 备注 | |
|
||||
|
||||
---
|
||||
|
||||
**版本:** 2026-07-30 v1
|
||||
**下次评审:** 本机 Phase 3(发布脚本 native 分支)合并 main 后
|
||||
Reference in New Issue
Block a user