refactor(msync): msync 整体搬迁至 SDK 并消除对 gvsds 的依赖

- msync 全部实现迁入 gvsdsdk(models/registry/ogm/transaction/auth +
  Django 集成层 middleware/sio/views/serializers/urls/admin/apps/migrations)
- 新增 gvsdsdk/secrets.py(secrets.ini 读取)与 serializer_base.py
  (PascalCase DRF 序列化器基类)
- msync 0001 迁移改用 gvsdsdk.models.UUIDField 并与模型 deep_deconstruct
  完全对齐(消除 makemigrations 漂移),dependencies 置空
- msync apps.py 移除 gvsds.app_config 软依赖,域名仅经环境变量/settings 解析
- 验证:屏蔽 gvsds 包后 50 个 gvsdsdk 子模块独立导入通过
This commit is contained in:
TermiNexus committed 2026-08-10 22:23:52 +08:00
1 parent fdac1d6e79
commit 53af9884df
22 files changed
+3975 -62

No files matched your search

+558
View File
@@ -0,0 +1,558 @@
# MSync — 微服务通信库
> GVsds 平台微服务架构的多端通信、服务间验权、对象全局映射(OGM)与分布式事务管理。
> 本服务器是**主服务器**(账本服务器),存储所有分布式服务地址,服务间需验权来互相获取 OGM 等信息。
> 实现位于 `gvsdsdk/msync`(SDK 原生实现,非封装层)。
---
## 概述
MSync 是为 GVsds 多域微服务平台设计的通信中间件,运行于**主服务器**(账本服务器),提供以下核心能力:
- **服务注册与发现** — 微服务节点的注册、注销、心跳、密钥轮换
- **服务间 HMAC 签名认证** — 基于 HMAC-SHA256 的双向验权,自包含令牌签发
- **OGM 对象全局映射** — 跨服务对象身份注册与解析
- **分布式事务管理** — 基于 Saga 模式的跨服务事务编排、补偿与重试
### 什么是"账本服务器"?
账本服务器即**主服务器**,它维护一张"账本"——记录所有分布式微服务的地址、状态、密钥等信息。任何微服务要加入集群,必须先向账本服务器注册;任何微服务要与其他服务通信,必须先通过账本服务器验权获取令牌。
---
## 架构
```
┌──────────────────────────────────┐
│ 账本服务器 (Master) │
│ │
│ ┌──────────┐ ┌──────────────┐ │
│ │ Registry │ │ Auth (HMAC) │ │
│ │ (服务账本)│ │ (服务间验权) │ │
│ └──────────┘ └──────────────┘ │
│ ┌──────────┐ ┌──────────────┐ │
│ │ OGM │ │ Transaction │ │
│ │(对象映射)│ │ (Saga编排) │ │
│ └──────────┘ └──────────────┘ │
└──────────┬───────────────────────┘
│
┌────────────────┼────────────────┐
│ │ │
┌─────┴─────┐ ┌────┴─────┐ ┌─────┴─────┐
│ GERP │ │ CRB │ │ T4 │
│ (企业ERP) │ │(货币经济)│ │(远程桌面) │
└───────────┘ └──────────┘ └───────────┘
```
---
## 认证流程
```
工作服务 账本服务器 目标服务
(Worker) (Master) (Target)
1. 构建 AuthPayload ──────▶
{service, challenge, ts}
2. HMAC-SHA256 签名 ──────▶
signature = HMAC(payload, secret)
3. 查找 ServiceNode
4. 验证签名
5. 签发 AccessToken + RefreshToken
◀──── 返回令牌对
6. 携带令牌请求 ─────────────────────────────────────────▶
Authorization: Service <AccessToken>
7. 本地验证令牌
8. 检查权限范围
9. 执行业务逻辑
◀───────────────────────────────── 返回结果
```
### 令牌机制
| 令牌类型 | 有效期 | 用途 |
|---------|--------|------|
| AccessToken | 2 小时 | 服务间请求认证 |
| RefreshToken | 30 天 | 刷新 AccessToken |
令牌为**自包含格式**(类似 JWT),目标服务可使用主密钥本地验证,无需回调账本服务器。
### 权限范围(Scope)
| Scope | 说明 |
|-------|------|
| `ogm:read` | 读取 OGM 映射 |
| `ogm:write` | 注册/更新 OGM 映射 |
| `registry:read` | 发现服务 |
| `registry:write` | 注册/注销服务 |
| `tx:read` | 查询事务状态 |
| `tx:write` | 创建/执行/补偿事务 |
| `*` | 全部权限 |
### 密钥认证强制
- 新服务注册必须携带 `ServiceSecret`(与 `ServiceSecretHash` 自洽),否则拒绝
- 除 register / auth / auth:refresh 外,所有 REST 与 SIO 业务操作必须持密钥签发的令牌
- 心跳/注销仅允许操作本服务
---
## 分布式事务管理
### 远程 + 本地双层架构
MSync 采用**远程事务管理器 + 本地事务管理器**双层架构:
```
主服务器 (远程事务管理器 TransactionManager)
│
│ 创建事务,编排步骤,管理状态
│
├──▶ POST {gerp_domain}/api/msync/local/execute/
│ {step_name: "create_user", payload: {...}}
│ → GERP 的 LocalTransactionManager.execute_step()
│
├──▶ POST {crb_domain}/api/msync/local/execute/
│ {step_name: "create_wallet", payload: {...}}
│ → CRB 的 LocalTransactionManager.execute_step()
│
└──▶ POST {crb_domain}/api/msync/local/compensate/
{step_name: "create_wallet", payload: {...}, action_result: {...}}
→ CRB 的 LocalTransactionManager.compensate_step()
```
- **远程事务管理器**(主服务器):创建事务、编排步骤、管理状态、调用各服务的本地端点
- **本地事务管理器**(每个工作服务):注册步骤处理器、执行本地逻辑、补偿本地操作
### Saga 模式
```
Step 0 (GERP) Step 1 (CRB) Step 2 (T4)
创建用户 ────────▶ 创建钱包 ────────▶ 注册终端
│ │ │
│ 补偿:删除用户 │ 补偿:删除钱包 │ 补偿:注销终端
◀──────────────── ◀────────────────── ◀
```
- **正向执行**:远程管理器按步骤顺序调用各服务的本地 execute 端点
- **补偿回滚**:任一步骤失败时,远程管理器逆序调用已完成步骤的本地 compensate 端点
- **超时检测**:自动标记超时事务并触发补偿
- **重试机制**:正向步骤失败可 `RetryStep`,补偿失败可 `RetryCompensate`(可配置最大次数)
### 事务生命周期
```
Pending ──▶ Running ──▶ Committed (全部步骤成功)
│
├─▶ Failed ──▶ Compensating ──▶ Compensated (补偿成功)
│
└─▶ TimedOut ──▶ Compensating ──▶ Compensated
```
| 状态 | 代码 | 说明 |
|------|------|------|
| 待执行 | 0 | 事务已创建,尚未开始 |
| 执行中 | 1 | 正在执行步骤 |
| 已提交 | 2 | 全部步骤成功完成 |
| 补偿中 | 3 | 正在逆序执行补偿操作 |
| 已补偿 | 4 | 补偿操作全部完成 |
| 失败 | 5 | 步骤执行失败且补偿未完成 |
| 超时 | 6 | 事务超时 |
### 步骤状态
| 状态 | 代码 | 说明 |
|------|------|------|
| 待执行 | 0 | 步骤尚未开始 |
| 执行中 | 1 | 正在调用 action 端点 |
| 已完成 | 2 | action 调用成功 |
| 补偿中 | 3 | 正在调用 compensate 端点 |
| 已补偿 | 4 | compensate 调用成功 |
| 失败 | 5 | 调用失败 |
| 已跳过 | 6 | 事务取消时跳过 |
---
## 数据模型
### ServiceNode — 服务节点(账本记录)
| 字段 | 类型 | 说明 |
|------|------|------|
| ServiceUUID | BINARY(16) PK | 服务唯一标识 |
| ServiceName | VARCHAR(128) UNIQUE | 服务名称 |
| ServiceDomain | VARCHAR(256) | 服务域名/URL |
| ServiceSecretHash | VARCHAR(128) | bcrypt 哈希密钥 |
| ServiceScope | JSON | 权限范围列表 |
| ServiceStatus | SMALLINT | 状态 (1=活跃) |
| ServiceType | SMALLINT | 类型 (0=工作节点, 1=主节点) |
| ServiceMetadata | JSON | 服务元数据 |
| CreateTime | DATETIME | 注册时间 |
| LastHeartbeat | DATETIME | 最后心跳时间 |
### ServiceToken — 服务令牌
| 字段 | 类型 | 说明 |
|------|------|------|
| TokenUUID | BINARY(16) PK | 令牌唯一标识 |
| ServiceUUID | FK → ServiceNode | 所属服务 |
| AccessToken | VARCHAR(512) UNIQUE | 访问令牌 |
| RefreshToken | VARCHAR(512) UNIQUE | 刷新令牌 |
| TokenScope | JSON | 令牌权限范围 |
| ExpiresAt | DATETIME | 访问令牌过期时间 |
| RefreshExpiresAt | DATETIME | 刷新令牌过期时间 |
| TokenStatus | SMALLINT | 状态 (1=有效, 0=已撤销) |
### ObjectGlobalMapping — OGM 对象全局映射
| 字段 | 类型 | 说明 |
|------|------|------|
| OGMUUID | BINARY(16) PK | 映射唯一标识 |
| ObjectType | VARCHAR(64) | 对象类型 (user, company, asset...) |
| ObjectLocalID | VARCHAR(128) | 本地 ID |
| ObjectGlobalID | VARCHAR(128) UNIQUE | 全局 ID |
| OwnerServiceUUID | FK → ServiceNode | 所属服务 |
| ObjectMetadata | JSON | 对象元数据 |
| ObjectStatus | SMALLINT | 状态 (1=活跃, 0=已注销) |
**唯一约束**: `(ObjectType, ObjectLocalID, OwnerServiceUUID)`
### DistributedTransaction — 分布式事务
| 字段 | 类型 | 说明 |
|------|------|------|
| TransactionUUID | BINARY(16) PK | 事务唯一标识 |
| TransactionCode | VARCHAR(64) UNIQUE | 事务代码 (DTX-XXXXXXXXXXXX) |
| TransactionType | VARCHAR(16) | 事务类型 (saga/2pc) |
| TransactionName | VARCHAR(128) | 事务名称 |
| TransactionDescription | VARCHAR(512) | 描述 |
| TransactionStatus | SMALLINT | 事务状态 |
| InitiatorServiceUUID | FK → ServiceNode | 发起服务 |
| ReferenceOGMUUID | BINARY(16) | 关联 OGM |
| TransactionPayload | JSON | 事务级载荷 |
| TransactionResult | JSON | 事务结果 |
| TimeoutSeconds | INT | 超时秒数 (默认 300) |
| CreateTime | DATETIME | 创建时间 |
| EndTime | DATETIME | 结束时间 |
### TransactionStep — 事务步骤
| 字段 | 类型 | 说明 |
|------|------|------|
| StepUUID | BINARY(16) PK | 步骤唯一标识 |
| TransactionUUID | FK → DistributedTransaction | 所属事务 |
| StepIndex | INT | 步骤序号 |
| StepName | VARCHAR(128) | 步骤名称 |
| ParticipantServiceUUID | FK → ServiceNode | 参与服务 |
| StepStatus | SMALLINT | 步骤状态 |
| ActionEndpoint | VARCHAR(512) | 正向操作端点 |
| ActionMethod | VARCHAR(8) | 正向操作方法 |
| ActionPayload | JSON | 正向操作载荷 |
| CompensateEndpoint | VARCHAR(512) | 补偿操作端点 |
| CompensateMethod | VARCHAR(8) | 补偿操作方法 |
| CompensatePayload | JSON | 补偿操作载荷 |
| StepResult | JSON | 步骤执行结果 |
| RetryCount | INT | 已重试次数 |
| MaxRetries | INT | 最大重试次数 (默认 3) |
---
## API 端点
### 认证
| 方法 | 端点 | 说明 |
|------|------|------|
| POST | `/api/msync/auth/token/` | 服务认证获取令牌 |
| POST | `/api/msync/auth/token/refresh/` | 刷新令牌 |
### 服务注册与发现(账本)
| 方法 | 端点 | 说明 |
|------|------|------|
| GET | `/api/msync/registry/` | 列出所有已注册服务 |
| POST | `/api/msync/registry/` | 注册新服务(必须携带密钥) |
| GET | `/api/msync/registry/<name>/` | 发现指定服务 |
| POST | `/api/msync/registry/<name>/heartbeat/` | 服务心跳(需令牌,仅本服务) |
| DELETE | `/api/msync/registry/<name>/deregister/` | 注销服务(需令牌,仅本服务) |
### OGM 对象全局映射
| 方法 | 端点 | 说明 |
|------|------|------|
| POST | `/api/msync/ogm/` | 注册对象映射 |
| POST | `/api/msync/ogm/batch/` | 批量注册 |
| GET | `/api/msync/ogm/<global_id>/` | 解析全局 ID |
| GET | `/api/msync/ogm/query/` | 按类型/服务查询 |
| PUT | `/api/msync/ogm/<global_id>/update/` | 更新元数据 |
| DELETE | `/api/msync/ogm/<global_id>/deregister/` | 注销映射 |
### 分布式事务
| 方法 | 端点 | 说明 |
|------|------|------|
| POST | `/api/msync/tx/` | 创建事务 |
| GET | `/api/msync/tx/query/?status=0` | 按状态查询事务 |
| POST | `/api/msync/tx/timeout-check/` | 检测超时事务 |
| GET | `/api/msync/tx/<code>/` | 事务详情 |
| POST | `/api/msync/tx/<code>/execute/` | 执行事务 |
| POST | `/api/msync/tx/<code>/compensate/` | 补偿事务 |
| POST | `/api/msync/tx/<code>/cancel/` | 取消事务 |
| POST | `/api/msync/tx/<code>/retry/` | 重试正向失败步骤 |
| POST | `/api/msync/tx/<code>/retry-compensate/` | 重试补偿失败步骤 |
### 本地事务管理器(每个工作服务)
| 方法 | 端点 | 说明 |
|------|------|------|
| POST | `/api/msync/local/execute/` | 执行本地步骤(由主服务器调用) |
| POST | `/api/msync/local/compensate/` | 补偿本地步骤(由主服务器调用) |
| GET | `/api/msync/local/handlers/` | 列出已注册的步骤处理器 |
**创建事务示例**:
```json
POST /api/msync/tx/
{
"name": "用户注册流程",
"description": "跨服务用户注册:创建用户→创建钱包→注册终端",
"transaction_type": "saga",
"timeout_seconds": 300,
"steps": [
{
"name": "创建用户",
"participant_service": "gerp",
"action_endpoint": "https://gerp.gvsds.com/api/users/",
"action_payload": {"username": "newuser"},
"compensate_endpoint": "https://gerp.gvsds.com/api/users/compensate/",
"compensate_payload": {"username": "newuser"}
},
{
"name": "创建钱包",
"participant_service": "crb",
"action_endpoint": "https://crb.gvsds.com/api/wallets/",
"action_payload": {"owner": "newuser"},
"compensate_endpoint": "https://crb.gvsds.com/api/wallets/compensate/",
"compensate_payload": {"owner": "newuser"}
},
{
"name": "注册终端",
"participant_service": "t4",
"action_endpoint": "https://t4.gvsds.com/api/terminals/",
"action_payload": {"user": "newuser"},
"compensate_endpoint": "https://t4.gvsds.com/api/terminals/compensate/",
"compensate_payload": {"user": "newuser"}
}
]
}
```
**执行事务**:
```json
POST /api/msync/tx/DTX-ABC123DEF456/execute/
{
"access_token": "eyJhbGci..."
}
```
**重试失败步骤**:
```json
POST /api/msync/tx/DTX-ABC123DEF456/retry/
{
"step_index": 1,
"access_token": "eyJhbGci..."
}
```
**重试补偿失败步骤**:
```json
POST /api/msync/tx/DTX-ABC123DEF456/retry-compensate/
{
"step_index": 1,
"access_token": "eyJhbGci..."
}
```
---
## Python API
### 服务注册(账本操作)
```python
from gvsdsdk.msync.registry import ServiceRegistry, RegistryError
node = ServiceRegistry.Register(
name='gerp',
domain='https://gerp.gvsds.com',
secret='my-super-secret-key',
scope=['ogm:read', 'ogm:write', 'tx:read', 'tx:write'],
)
info = ServiceRegistry.Discover('gerp')
ServiceRegistry.Heartbeat('gerp')
ServiceRegistry.RotateSecret('gerp', 'new-secret-key')
ServiceRegistry.Deregister('gerp')
```
### 服务客户端
```python
from gvsdsdk.msync.client import ServiceClient
client = ServiceClient('gerp', 'secret', 'https://api.gvsds.com')
client.Authenticate()
ogm = client.ResolveOGM('user:550e8400')
info = client.DiscoverService('crb')
```
### OGM 管理
```python
from gvsdsdk.msync.ogm import OGMManager
ogm = OGMManager.Register('user', 'uuid-1', 'core', global_id='user:uuid1')
info = OGMManager.Resolve('user:uuid1')
users = OGMManager.QueryByType('user')
OGMManager.Deregister('user:uuid1')
```
### 分布式事务
```python
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
# 创建事务
dtx = TransactionManager.Create(
name='用户注册流程',
steps_data=[
{
'name': 'create_user',
'participant_service': 'gerp',
'action_payload': {'username': 'newuser'},
'compensate_payload': {'username': 'newuser'},
},
{
'name': 'create_wallet',
'participant_service': 'crb',
'action_payload': {'owner': 'newuser'},
'compensate_payload': {'owner': 'newuser'},
},
],
timeout_seconds=300,
)
# 执行事务(远程管理器自动调用各服务的本地管理器)
dtx = TransactionManager.Execute(dtx.TransactionCode, access_token='...')
# 查看详情
detail = TransactionManager.GetDetail(dtx.TransactionCode)
# 手动补偿
dtx = TransactionManager.Compensate(dtx.TransactionCode, access_token='...')
# 重试正向失败步骤
step = TransactionManager.RetryStep(dtx.TransactionCode, step_index=1, access_token='...')
# 重试补偿失败步骤
step = TransactionManager.RetryCompensate(dtx.TransactionCode, step_index=1, access_token='...')
# 取消待执行事务
TransactionManager.Cancel(dtx.TransactionCode)
# 检测超时
timed_out = TransactionManager.CheckTimeouts()
# 按状态查询
running = TransactionManager.QueryByStatus(1)
```
### 本地事务管理器(工作服务端)
```python
from gvsdsdk.msync.local_tx import local_tx
# 注册步骤处理器(在服务启动时)
def create_user_action(payload):
user = User.objects.create(username=payload['username'])
return {'user_uuid': str(user.UserUUID)}
def create_user_compensate(payload, action_result):
User.objects.filter(username=payload['username']).delete()
return {'deleted': True}
local_tx.register(
'create_user',
action=create_user_action,
compensate=create_user_compensate,
)
# 列出已注册处理器
handlers = local_tx.list_handlers()
# 注销处理器
local_tx.unregister('create_user')
```
### 权限装饰器
```python
from gvsdsdk.msync.middleware import require_service_scope, ServicePermission
@require_service_scope('ogm:read')
def my_view(request):
pass
class MyViewSet(ViewSet):
permission_classes = [ServicePermission('tx:read', 'tx:write')]
```
---
## 配置
### settings.py
```python
INSTALLED_APPS = [
'gvsdsdk.msync',
]
MIDDLEWARE = [
'gvsdsdk.msync.middleware.ServiceAuthMiddleware',
]
```
### 数据库路由
msync 使用 `default` 数据库(Users 库),与 core、gerp 共享。
### secrets.ini
```ini
[msync]
MASTER_SECRET = your-master-secret-for-token-signing
```
---
## 文件清单
| 文件 | 说明 |
|------|------|
| `models.py` | 5 个数据模型(ServiceNode, ServiceToken, ObjectGlobalMapping, DistributedTransaction, TransactionStep) |
| `auth.py` | HMAC 签名 + 自包含令牌编解码 + 服务端签发/验证 |
| `client.py` | ServiceClient 跨服务请求客户端 |
| `middleware.py` | 服务认证中间件 + 权限装饰器 |
| `registry.py` | 服务注册/发现/心跳/密钥轮换 |
| `ogm.py` | OGM 对象全局映射管理 |
| `transaction.py` | 远程事务管理器(Saga 编排、补偿、重试、超时,调用本地管理器) |
| `local_tx.py` | 本地事务管理器(步骤处理器注册、本地执行与补偿) |
| `sio.py` | Socket.IO 服务端(注册/认证/心跳/OGM/事务事件 + 密钥认证守卫) |
| `views.py` | REST API 视图 |
| `serializers.py` | DRF 序列化器 |
| `urls.py` | API 路由 |
| `admin.py` | Django Admin 注册 |
+17 -19
View File
@@ -1,19 +1,17 @@
"""GVSDS SDK MSYNC 子包
将 MSYNC 相关模块统一组织在 gvsdsdk.msync 命名空间下。
自有实现(无 Django 依赖,子服务器可直接使用):
- auth: HMAC-SHA256 签名验证、服务令牌编解码
MSYNC 全部实现位于 gvsdsdk.msync(原生实现,非封装层):
- models: ServiceNode / ServiceToken / OGM / 分布式事务模型
- auth: HMAC-SHA256 签名验证、服务令牌编解码、服务端签发/验证
- registry / ogm / transaction: 服务注册发现 / OGM 管理 / Saga 编排
- client: 微服务通信客户端(REST)
- local_tx: 本地事务步骤处理器
服务端转发(依赖 Django ORM,仅在主服务器可用):
- registry: 服务注册/发现
- ogm: 对象全局映射
- transaction: Saga 分布式事务
- middleware / views / serializers / urls / sio / admin: Django 集成层
"""
# 自有实现(无 Django 依赖,始终可用)
default_app_config = 'gvsdsdk.msync.apps.MsyncConfig'
# 自有实现(无 Django ORM 依赖,子服务器可直接使用)
from gvsdsdk.msync.auth import ( # noqa
SignPayload,
VerifySignature,
@@ -26,7 +24,7 @@ from gvsdsdk.msync.auth import ( # noqa
from gvsdsdk.msync.client import ServiceClient, ServiceClientError # noqa
from gvsdsdk.msync.local_tx import LocalTransactionManager, StepHandler, local_tx # noqa
# 服务端转发(Django ORM 依赖,延迟导入避免子服务器初始化报错)
# ORM 依赖的模块延迟导入(仅在主服务器使用)
def __getattr__(name):
_server_exports = {
'ServiceRegistry': 'gvsdsdk.msync.registry',
@@ -36,10 +34,10 @@ def __getattr__(name):
'TransactionManager': 'gvsdsdk.msync.transaction',
'TransactionError': 'gvsdsdk.msync.transaction',
'StepExecutionError': 'gvsdsdk.msync.transaction',
'AuthenticateService': 'gvsds.apps.msync.auth',
'RefreshServiceToken': 'gvsds.apps.msync.auth',
'ValidateAccessToken': 'gvsds.apps.msync.auth',
'IssueServiceToken': 'gvsds.apps.msync.auth',
'AuthenticateService': 'gvsdsdk.msync.auth',
'RefreshServiceToken': 'gvsdsdk.msync.auth',
'ValidateAccessToken': 'gvsdsdk.msync.auth',
'IssueServiceToken': 'gvsdsdk.msync.auth',
}
if name in _server_exports:
import importlib
@@ -57,7 +55,7 @@ __all__ = [
'SignAuthPayload',
'VerifyAuthPayload',
'ServiceTokenEncoder',
# auth (服务端转发)
# auth (服务端)
'AuthenticateService',
'RefreshServiceToken',
'ValidateAccessToken',
@@ -69,13 +67,13 @@ __all__ = [
'LocalTransactionManager',
'StepHandler',
'local_tx',
# registry (服务端转发)
# registry (服务端)
'ServiceRegistry',
'RegistryError',
# ogm (服务端转发)
# ogm (服务端)
'OGMManager',
'OGMError',
# transaction (服务端转发)
# transaction (服务端)
'TransactionManager',
'TransactionError',
'StepExecutionError',
+45
View File
@@ -0,0 +1,45 @@
from django.contrib import admin
from gvsdsdk.msync.models import (
ServiceNode,
ServiceToken,
ObjectGlobalMapping,
DistributedTransaction,
TransactionStep,
)
@admin.register(ServiceNode)
class ServiceNodeAdmin(admin.ModelAdmin):
list_display = ('ServiceName', 'ServiceDomain', 'ServiceType', 'ServiceStatus', 'CreateTime', 'LastHeartbeat')
list_filter = ('ServiceType', 'ServiceStatus')
search_fields = ('ServiceName', 'ServiceDomain')
readonly_fields = ('ServiceUUID', 'ServiceSecretHash', 'CreateTime')
@admin.register(ServiceToken)
class ServiceTokenAdmin(admin.ModelAdmin):
list_display = ('TokenUUID', 'ServiceUUID', 'ExpiresAt', 'TokenStatus', 'CreateTime')
list_filter = ('TokenStatus',)
readonly_fields = ('TokenUUID', 'AccessToken', 'RefreshToken', 'CreateTime')
@admin.register(ObjectGlobalMapping)
class ObjectGlobalMappingAdmin(admin.ModelAdmin):
list_display = ('ObjectGlobalID', 'ObjectType', 'ObjectLocalID', 'OwnerServiceUUID', 'ObjectStatus', 'CreateTime')
list_filter = ('ObjectType', 'ObjectStatus')
search_fields = ('ObjectGlobalID', 'ObjectLocalID')
@admin.register(DistributedTransaction)
class DistributedTransactionAdmin(admin.ModelAdmin):
list_display = ('TransactionCode', 'TransactionName', 'TransactionType', 'TransactionStatus', 'InitiatorServiceUUID', 'CreateTime', 'EndTime')
list_filter = ('TransactionType', 'TransactionStatus')
search_fields = ('TransactionCode', 'TransactionName')
readonly_fields = ('TransactionUUID', 'CreateTime')
@admin.register(TransactionStep)
class TransactionStepAdmin(admin.ModelAdmin):
list_display = ('StepUUID', 'TransactionUUID', 'StepIndex', 'StepName', 'StepStatus', 'ParticipantServiceUUID', 'RetryCount')
list_filter = ('StepStatus',)
readonly_fields = ('StepUUID', 'CreateTime')
+68
View File
@@ -0,0 +1,68 @@
from django.apps import AppConfig
class MsyncConfig(AppConfig):
default_auto_field = 'django.db.models.BigAutoField'
name = 'gvsdsdk.msync'
verbose_name = '微服务通信'
verbose_name_plural = '微服务通信'
def ready(self):
# 仅在主进程启动时确保 gvsds-master 节点存在
import os
if os.environ.get('RUN_MAIN') == 'true' or not os.environ.get('RUN_MAIN'):
try:
self._ensure_master_node()
except Exception:
pass # 数据库可能尚未迁移
def _ensure_master_node(self):
from gvsdsdk.msync.models import ServiceNode, _utcnow
import uuid
service_domain = self._resolve_service_domain()
master_uuid = uuid.UUID('00000000-0000-0000-0000-000000000001')
try:
node = ServiceNode.objects.get(ServiceName='gvsds-master')
# 更新心跳、状态和元数据
node.ServiceStatus = 1
node.ServiceScope = ['*']
node.ServiceDomain = service_domain
node.ServiceType = 1
node.ServiceMetadata = {'DBName': 'Users'}
node.LastHeartbeat = _utcnow()
node.save(update_fields=['ServiceStatus', 'ServiceScope', 'ServiceDomain', 'ServiceType', 'ServiceMetadata', 'LastHeartbeat'])
except ServiceNode.DoesNotExist:
import bcrypt
secret = 'gvsds-master-internal'
secret_hash = bcrypt.hashpw(secret.encode('utf-8'), bcrypt.gensalt()).decode('utf-8')
node = ServiceNode(
ServiceUUID=master_uuid.bytes,
ServiceName='gvsds-master',
ServiceDomain=service_domain,
ServiceSecretHash=secret_hash,
ServiceScope=['*'],
ServiceStatus=1,
ServiceType=1,
ServiceMetadata={'DBName': 'Users'},
LastHeartbeat=_utcnow(),
)
node.SetSecret(secret)
node.save()
@staticmethod
def _resolve_service_domain():
"""解析主服务器公开域名:环境变量 → settings → 默认值(不依赖 gvsds)"""
import os
domain = os.environ.get('MSYNC_SERVICE_DOMAIN') or os.environ.get('APP_MSYNC_SERVICE_DOMAIN')
if domain:
return domain
try:
from django.conf import settings as _settings
domain = getattr(_settings, 'GVSDSDK_MSYNC_MASTER_URL', None)
if domain:
return str(domain).rstrip('/')
except Exception:
pass
return 'https://www.gvsds.com'
+154 -3
View File
@@ -1,8 +1,8 @@
"""MSYNC 认证模块
提供 HMAC-SHA256 签名验证、服务令牌编解码等纯函数。
Django ORM 依赖的函数(AuthenticateService, IssueServiceToken 等)
仍在 gvsds.apps.msync.auth 中,由服务端调用。
提供 HMAC-SHA256 签名验证、服务令牌编解码等纯函数,
以及 ORM 依赖的服务端认证函数(AuthenticateService, IssueServiceToken 等,
模型延迟导入,主服务器侧使用)。
"""
import hmac
@@ -13,6 +13,7 @@ import base64
import binascii
import uuid
import logging
import traceback
import datetime
logger = logging.getLogger(__name__)
@@ -140,3 +141,153 @@ class ServiceTokenEncoder:
return Payload
except (ValueError, TypeError, json.JSONDecodeError, binascii.Error):
return None
# ─── ORM 依赖的认证函数(主服务器侧使用,模型延迟导入)──────────────
def _get_master_secret():
"""获取 MASTER_SECRET:环境变量 MSYNC_MASTER_SECRET 优先,其次 secrets.ini"""
from gvsdsdk.secrets import get as _secret_get
secret = _secret_get('msync', 'MASTER_SECRET', fallback=None)
if not secret:
from django.core.exceptions import ImproperlyConfigured
raise ImproperlyConfigured('msync MASTER_SECRET 未配置,请在 secrets.ini 或环境变量 MSYNC_MASTER_SECRET 中设置')
return secret
def IssueServiceToken(service_node, access_lifetime_hours=2, refresh_lifetime_days=30):
"""为服务节点签发令牌对(AccessToken + RefreshToken)"""
from gvsdsdk.msync.models import ServiceToken
master_secret = _get_master_secret()
now_aware = datetime.datetime.now(datetime.timezone.utc)
access_expires_aware = now_aware + datetime.timedelta(hours=access_lifetime_hours)
refresh_expires_aware = now_aware + datetime.timedelta(days=refresh_lifetime_days)
now_naive = now_aware.replace(tzinfo=None)
access_expires = access_expires_aware.replace(tzinfo=None)
refresh_expires = refresh_expires_aware.replace(tzinfo=None)
service_scope = service_node.ServiceScope
if service_scope is None:
service_scope = ['ogm:read', 'ogm:write', 'registry:read']
elif isinstance(service_scope, str):
service_scope = [service_scope]
encoder = ServiceTokenEncoder(master_secret)
access_token = encoder.Encode(
service_name=service_node.ServiceName,
scope=service_scope,
expires_at=access_expires_aware,
)
refresh_token = encoder.Encode(
service_name=service_node.ServiceName,
scope=service_scope,
expires_at=refresh_expires_aware,
)
import uuid
token = ServiceToken(
TokenUUID=uuid.uuid4().bytes,
ServiceUUID=service_node,
AccessToken=access_token,
RefreshToken=refresh_token,
TokenScope=service_scope,
ExpiresAt=access_expires,
RefreshExpiresAt=refresh_expires,
)
token.save()
return token
def AuthenticateService(service_name, signed_payload, signature):
"""验证服务身份并签发令牌"""
from gvsdsdk.msync.models import ServiceNode, ServiceToken
try:
node = ServiceNode.objects.get(ServiceName=service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
logger.error(f'[AUTH] 服务节点不存在或未激活: {service_name}')
return None
except Exception as e:
logger.error(f'查询服务节点异常 [{service_name}]: {e}\n{traceback.format_exc()}')
return None
import time
ts = signed_payload.get('ts', 0)
now = time.time()
ts_diff = abs(now - ts)
if ts_diff > 300:
logger.error(f'[AUTH] 时间戳校验失败: ts={ts}, now={now}, diff={ts_diff:.1f}s')
return None
raw_secret = node.GetSecret()
if raw_secret is None:
logger.error(f'服务 {service_name} 密钥加密副本缺失,请重新注册')
return None
if not VerifyAuthPayload(signed_payload, signature, raw_secret):
logger.error(f'[AUTH] 签名验证失败: service={service_name}, payload={signed_payload}')
return None
try:
ServiceToken.objects.filter(ServiceUUID=node, TokenStatus=1).update(TokenStatus=0)
except Exception as e:
logger.error(f'撤销旧令牌异常 [{service_name}]: {e}')
try:
return IssueServiceToken(node)
except Exception as e:
logger.error(f'签发服务令牌异常 [{service_name}]: {e}\n{traceback.format_exc()}')
return None
def ValidateAccessToken(access_token):
"""验证 AccessToken 并返回载荷信息"""
from gvsdsdk.msync.models import ServiceToken
master_secret = _get_master_secret()
encoder = ServiceTokenEncoder(master_secret)
payload = encoder.Decode(access_token)
if payload is None:
return None
try:
token_record = ServiceToken.objects.get(AccessToken=access_token, TokenStatus=1)
if token_record.IsExpired():
token_record.Revoke()
return None
except ServiceToken.DoesNotExist:
return None
return payload
def RefreshServiceToken(refresh_token):
"""使用 RefreshToken 刷新令牌对"""
from gvsdsdk.msync.models import ServiceToken, ServiceNode
master_secret = _get_master_secret()
encoder = ServiceTokenEncoder(master_secret)
payload = encoder.Decode(refresh_token)
if payload is None:
return None
try:
old_token = ServiceToken.objects.get(RefreshToken=refresh_token, TokenStatus=1)
except ServiceToken.DoesNotExist:
return None
if old_token.IsRefreshExpired():
old_token.Revoke()
return None
service_name = payload.get('sub')
try:
node = ServiceNode.objects.get(ServiceName=service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
return None
old_token.Revoke()
return IssueServiceToken(node)
+154
View File
@@ -0,0 +1,154 @@
"""模块:服务请求验证中间件与权限装饰器
用于验证来自其他微服务的请求,提取服务身份信息。
在 Django 视图中使用 @require_service_scope 装饰器保护端点。"""
import functools
import logging
from django.http import JsonResponse
from rest_framework.response import Response
from gvsdsdk.msync.auth import ValidateAccessToken
logger = logging.getLogger(__name__)
class ServiceAuthMiddleware:
"""服务认证中间件——从请求中提取并验证服务令牌
将验证结果附加到 request 对象:
- request.service_name: 发起请求的服务名称
- request.service_scope: 服务的权限范围列表
- request.service_token: 原始令牌字符串
"""
SERVICE_AUTH_HEADER = 'Authorization'
SERVICE_AUTH_PREFIX = 'Service'
def __init__(self, get_response):
self.get_response = get_response
def __call__(self, request):
request.service_name = None
request.service_scope = []
request.service_token = None
auth_header = request.META.get('HTTP_AUTHORIZATION', '')
if auth_header.startswith(f'{self.SERVICE_AUTH_PREFIX} '):
token = auth_header[len(self.SERVICE_AUTH_PREFIX) + 1:]
payload = ValidateAccessToken(token)
if payload:
request.service_name = payload.get('sub')
request.service_scope = payload.get('scope', [])
request.service_token = token
response = self.get_response(request)
return response
def require_service_scope(*required_scopes):
"""视图装饰器——要求请求来自具有指定权限范围的已认证服务
用法:
@require_service_scope('ogm:read')
def my_view(request):
# request.service_name 已设置
...
"""
def decorator(view_func):
@functools.wraps(view_func)
def wrapper(request, *args, **kwargs):
if not request.service_name:
return JsonResponse(
{'code': 401, 'msg': '需要服务认证', 'data': None},
status=401,
)
if '*' not in request.service_scope:
for scope in required_scopes:
if scope not in request.service_scope:
return JsonResponse(
{'code': 403, 'msg': f'缺少权限: {scope}', 'data': None},
status=403,
)
return view_func(request, *args, **kwargs)
return wrapper
return decorator
def require_service_scope_drq(*required_scopes):
"""DRF 视图装饰器——返回 DRF Response 而非 Django JsonResponse"""
def decorator(view_func):
@functools.wraps(view_func)
def wrapper(self, request, *args, **kwargs):
if not request.service_name:
return Response(
{'code': 401, 'msg': '需要服务认证', 'data': None},
status=401,
)
if '*' not in request.service_scope:
for scope in required_scopes:
if scope not in request.service_scope:
return Response(
{'code': 403, 'msg': f'缺少权限: {scope}', 'data': None},
status=403,
)
return view_func(self, request, *args, **kwargs)
return wrapper
return decorator
class ServicePermission:
"""DRF 权限类——用于 ViewSet 的 permission_classes"""
def __init__(self, *required_scopes):
self.required_scopes = required_scopes
def has_permission(self, request, view):
if not request.service_name:
return False
if '*' in request.service_scope:
return True
return all(s in request.service_scope for s in self.required_scopes)
class ServiceTokenAuthentication:
"""DRF 认证类——从请求头中提取并验证服务令牌
配合 ServiceAuthMiddleware 使用,中间件已将验证结果附加到 request 对象。
此类使 DRF 的 request.user 和 request.auth 可用于权限检查。
"""
def authenticate(self, request):
service_name = getattr(request, 'service_name', None)
if not service_name:
return None
# 返回 (user, auth) 元组,使 DRF 认为请求已认证
# service_name 作为 user 标识,service_token 作为 auth
return (ServicePrincipal(request), request.service_token)
def authenticate_header(self, request):
return 'Service'
class ServicePrincipal:
"""服务主体——作为 DRF 认证后的 request.user"""
def __init__(self, request):
self.service_name = request.service_name
self.service_scope = request.service_scope
self.service_token = request.service_token
@property
def is_authenticated(self):
return True
def __str__(self):
return f'Service:{self.service_name}'
+151
View File
@@ -0,0 +1,151 @@
import datetime
from django.db import migrations, models
import django.db.models.deletion
import gvsdsdk.models
class Migration(migrations.Migration):
initial = True
dependencies = []
operations = [
migrations.CreateModel(
name='ServiceNode',
fields=[
('ServiceUUID', gvsdsdk.models.UUIDField(verbose_name='服务UUID', editable=False, primary_key=True, serialize=False, binary_length=16)),
('ServiceName', models.CharField(db_column='ServiceName', max_length=128, unique=True)),
('ServiceDomain', models.CharField(db_column='ServiceDomain', max_length=256)),
('ServiceSecretHash', models.CharField(db_column='ServiceSecretHash', max_length=128)),
('ServiceScope', models.JSONField(db_column='ServiceScope', default=list)),
('ServiceStatus', models.SmallIntegerField(db_column='ServiceStatus', default=1)),
('ServiceType', models.SmallIntegerField(db_column='ServiceType', default=0)),
('ServiceMetadata', models.JSONField(db_column='ServiceMetadata', default=dict)),
('CreateTime', models.DateTimeField(db_column='CreateTime', default=datetime.datetime.utcnow)),
('LastHeartbeat', models.DateTimeField(blank=True, db_column='LastHeartbeat', null=True)),
],
options={
'verbose_name': '服务节点',
'verbose_name_plural': '服务节点',
'db_table': 'ServiceNode',
'ordering': ['-CreateTime'],
},
),
migrations.CreateModel(
name='ServiceToken',
fields=[
('TokenUUID', gvsdsdk.models.UUIDField(verbose_name='令牌UUID', editable=False, primary_key=True, serialize=False, binary_length=16)),
('AccessToken', models.CharField(db_column='AccessToken', max_length=512, unique=True)),
('RefreshToken', models.CharField(db_column='RefreshToken', max_length=512, unique=True)),
('TokenScope', models.JSONField(db_column='TokenScope', default=list)),
('ExpiresAt', models.DateTimeField(db_column='ExpiresAt')),
('RefreshExpiresAt', models.DateTimeField(db_column='RefreshExpiresAt')),
('CreateTime', models.DateTimeField(db_column='CreateTime', default=datetime.datetime.utcnow)),
('TokenStatus', models.SmallIntegerField(db_column='TokenStatus', default=1)),
('ServiceUUID', models.ForeignKey(db_column='ServiceUUID', on_delete=django.db.models.deletion.CASCADE, related_name='tokens', to='msync.servicenode')),
],
options={
'verbose_name': '服务令牌',
'verbose_name_plural': '服务令牌',
'db_table': 'ServiceToken',
'ordering': ['-CreateTime'],
},
),
migrations.CreateModel(
name='ObjectGlobalMapping',
fields=[
('OGMUUID', gvsdsdk.models.UUIDField(verbose_name='OGM UUID', editable=False, primary_key=True, serialize=False, binary_length=16)),
('ObjectType', models.CharField(db_column='ObjectType', max_length=64)),
('ObjectLocalID', models.CharField(db_column='ObjectLocalID', max_length=128)),
('ObjectGlobalID', models.CharField(db_column='ObjectGlobalID', max_length=128, unique=True)),
('ObjectMetadata', models.JSONField(db_column='ObjectMetadata', default=dict)),
('ObjectStatus', models.SmallIntegerField(db_column='ObjectStatus', default=1)),
('CreateTime', models.DateTimeField(db_column='CreateTime', default=datetime.datetime.utcnow)),
('UpdateTime', models.DateTimeField(blank=True, db_column='UpdateTime', null=True)),
('OwnerServiceUUID', models.ForeignKey(db_column='OwnerServiceUUID', on_delete=django.db.models.deletion.CASCADE, related_name='owned_objects', to='msync.servicenode')),
],
options={
'verbose_name': '对象全局映射',
'verbose_name_plural': '对象全局映射',
'db_table': 'OGM',
'ordering': ['-CreateTime'],
'unique_together': {('ObjectType', 'ObjectLocalID', 'OwnerServiceUUID')},
},
),
migrations.AddIndex(
model_name='objectglobalmapping',
index=models.Index(fields=['ObjectType'], name='idx_ogm_type'),
),
migrations.AddIndex(
model_name='objectglobalmapping',
index=models.Index(fields=['ObjectGlobalID'], name='idx_ogm_gid'),
),
migrations.CreateModel(
name='DistributedTransaction',
fields=[
('TransactionUUID', gvsdsdk.models.UUIDField(verbose_name='事务UUID', editable=False, primary_key=True, serialize=False, binary_length=16)),
('TransactionCode', models.CharField(db_column='TransactionCode', max_length=64, unique=True)),
('TransactionType', models.CharField(choices=[('saga', 'Saga'), ('2pc', '两阶段提交')], db_column='TransactionType', default='saga', max_length=16)),
('TransactionName', models.CharField(db_column='TransactionName', max_length=128)),
('TransactionDescription', models.CharField(db_column='TransactionDescription', default='', max_length=512)),
('TransactionStatus', models.SmallIntegerField(choices=[(0, '待执行'), (1, '执行中'), (2, '已提交'), (3, '补偿中'), (4, '已补偿'), (5, '失败'), (6, '超时')], db_column='TransactionStatus', default=0)),
('ReferenceOGMUUID', gvsdsdk.models.UUIDField(verbose_name='关联OGM UUID', blank=True, null=True, binary_length=16)),
('TransactionPayload', models.JSONField(db_column='TransactionPayload', default=dict)),
('TransactionResult', models.JSONField(db_column='TransactionResult', default=dict)),
('TimeoutSeconds', models.IntegerField(db_column='TimeoutSeconds', default=300)),
('CreateTime', models.DateTimeField(db_column='CreateTime', default=datetime.datetime.utcnow)),
('UpdateTime', models.DateTimeField(blank=True, db_column='UpdateTime', null=True)),
('EndTime', models.DateTimeField(blank=True, db_column='EndTime', null=True)),
('InitiatorServiceUUID', models.ForeignKey(blank=True, db_column='InitiatorServiceUUID', null=True, on_delete=django.db.models.deletion.SET_NULL, related_name='initiated_dtx', to='msync.servicenode')),
],
options={
'verbose_name': '分布式事务',
'verbose_name_plural': '分布式事务',
'db_table': 'DistributedTransaction',
'ordering': ['-CreateTime'],
},
),
migrations.AddIndex(
model_name='distributedtransaction',
index=models.Index(fields=['TransactionStatus'], name='idx_dtx_status'),
),
migrations.AddIndex(
model_name='distributedtransaction',
index=models.Index(fields=['InitiatorServiceUUID'], name='idx_dtx_initiator'),
),
migrations.CreateModel(
name='TransactionStep',
fields=[
('StepUUID', gvsdsdk.models.UUIDField(verbose_name='步骤UUID', editable=False, primary_key=True, serialize=False, binary_length=16)),
('StepIndex', models.IntegerField(db_column='StepIndex')),
('StepName', models.CharField(db_column='StepName', max_length=128)),
('StepDescription', models.CharField(db_column='StepDescription', default='', max_length=512)),
('StepStatus', models.SmallIntegerField(choices=[(0, '待执行'), (1, '执行中'), (2, '已完成'), (3, '补偿中'), (4, '已补偿'), (5, '失败'), (6, '已跳过')], db_column='StepStatus', default=0)),
('ActionEndpoint', models.CharField(db_column='ActionEndpoint', max_length=512)),
('ActionMethod', models.CharField(db_column='ActionMethod', default='POST', max_length=8)),
('ActionPayload', models.JSONField(db_column='ActionPayload', default=dict)),
('CompensateEndpoint', models.CharField(db_column='CompensateEndpoint', default='', max_length=512)),
('CompensateMethod', models.CharField(db_column='CompensateMethod', default='POST', max_length=8)),
('CompensatePayload', models.JSONField(db_column='CompensatePayload', default=dict)),
('StepResult', models.JSONField(db_column='StepResult', default=dict)),
('RetryCount', models.IntegerField(db_column='RetryCount', default=0)),
('MaxRetries', models.IntegerField(db_column='MaxRetries', default=3)),
('CreateTime', models.DateTimeField(db_column='CreateTime', default=datetime.datetime.utcnow)),
('StartTime', models.DateTimeField(blank=True, db_column='StartTime', null=True)),
('EndTime', models.DateTimeField(blank=True, db_column='EndTime', null=True)),
('ParticipantServiceUUID', models.ForeignKey(blank=True, db_column='ParticipantServiceUUID', null=True, on_delete=django.db.models.deletion.SET_NULL, related_name='participated_steps', to='msync.servicenode')),
('TransactionUUID', models.ForeignKey(db_column='TransactionUUID', on_delete=django.db.models.deletion.CASCADE, related_name='steps', to='msync.distributedtransaction')),
],
options={
'verbose_name': '事务步骤',
'verbose_name_plural': '事务步骤',
'db_table': 'TransactionStep',
'ordering': ['TransactionUUID', 'StepIndex'],
},
),
migrations.AddIndex(
model_name='transactionstep',
index=models.Index(fields=['TransactionUUID', 'StepIndex'], name='idx_step_tx_order'),
),
]
@@ -0,0 +1,16 @@
from django.db import migrations, models
class Migration(migrations.Migration):
dependencies = [
('msync', '0001_initial'),
]
operations = [
migrations.AddField(
model_name='servicenode',
name='ServiceSecretEncrypted',
field=models.CharField(db_column='ServiceSecretEncrypted', default='', max_length=512),
),
]
+1
View File
@@ -0,0 +1 @@
# -*- coding: utf-8 -*-
+317
View File
@@ -0,0 +1,317 @@
"""模块:MSync 数据模型
微服务通信核心表:服务节点、服务令牌、对象全局映射(OGM)、分布式事务。
所有 ORM 字段与方法遵循帕斯卡命名法。
模型归属 app_label='msync',与 gvsds 主服务器 7 库路由及既有迁移兼容。"""
import uuid
import datetime
from django.db import models
from gvsdsdk.model_base import QModel
from gvsdsdk.models import UUIDField
def _utcnow():
"""返回 naive UTC datetime(兼容 Django USE_TZ=False)"""
return datetime.datetime.now(datetime.timezone.utc).replace(tzinfo=None)
class ServiceNode(QModel):
"""服务节点表——注册的微服务实例(账本服务器存储的所有分布式服务地址)"""
ServiceUUID = UUIDField(primary_key=True, verbose_name='服务UUID')
ServiceName = models.CharField(max_length=128, unique=True, verbose_name='服务名称')
ServiceDomain = models.CharField(max_length=256, verbose_name='服务域名')
ServiceSecretHash = models.CharField(max_length=128, verbose_name='密钥哈希')
ServiceSecretEncrypted = models.CharField(max_length=512, default='', verbose_name='加密密钥')
ServiceScope = models.JSONField(default=list, verbose_name='服务权限')
ServiceStatus = models.SmallIntegerField(default=1, verbose_name='服务状态')
ServiceType = models.SmallIntegerField(default=0, verbose_name='服务类型')
ServiceMetadata = models.JSONField(default=dict, verbose_name='服务元数据')
CreateTime = models.DateTimeField(default=_utcnow, verbose_name='创建时间')
LastHeartbeat = models.DateTimeField(null=True, blank=True, verbose_name='最后心跳')
class Meta:
db_table = 'ServiceNode'
app_label = 'msync'
verbose_name = '服务节点'
verbose_name_plural = '服务节点'
ordering = ['-CreateTime']
def __str__(self):
return self.ServiceName
@property
def IsMaster(self):
return self.ServiceType == 1
@property
def IsActive(self):
return self.ServiceStatus == 1
def VerifySecret(self, raw_secret):
import bcrypt
try:
return bcrypt.checkpw(
raw_secret.encode('utf-8'),
self.ServiceSecretHash.encode('utf-8')
)
except Exception:
return False
def SetSecret(self, raw_secret):
from gvsdsdk.auth.crypto import SymmetricEncrypt
from gvsdsdk.secrets import get as _secret_get
master = _secret_get('msync', 'MASTER_SECRET', fallback='gvsds-msync-default-secret')
self.ServiceSecretEncrypted = SymmetricEncrypt(raw_secret, master)
def GetSecret(self):
if not self.ServiceSecretEncrypted:
return None
from gvsdsdk.auth.crypto import SymmetricDecrypt
from gvsdsdk.secrets import get as _secret_get
master = _secret_get('msync', 'MASTER_SECRET', fallback='gvsds-msync-default-secret')
try:
return SymmetricDecrypt(self.ServiceSecretEncrypted, master)
except Exception:
return None
def HasScope(self, scope_name):
if not self.ServiceScope:
return False
return scope_name in self.ServiceScope or '*' in self.ServiceScope
class ServiceToken(QModel):
"""服务令牌表——服务间认证凭据"""
TokenUUID = UUIDField(primary_key=True, verbose_name='令牌UUID')
ServiceUUID = models.ForeignKey(
ServiceNode, on_delete=models.CASCADE,
related_name='tokens', verbose_name='服务UUID',
)
AccessToken = models.CharField(max_length=512, unique=True, verbose_name='访问令牌')
RefreshToken = models.CharField(max_length=512, unique=True, verbose_name='刷新令牌')
TokenScope = models.JSONField(default=list, verbose_name='令牌权限')
ExpiresAt = models.DateTimeField(verbose_name='过期时间')
RefreshExpiresAt = models.DateTimeField(verbose_name='刷新过期时间')
CreateTime = models.DateTimeField(default=_utcnow, verbose_name='创建时间')
TokenStatus = models.SmallIntegerField(default=1, verbose_name='令牌状态')
class Meta:
db_table = 'ServiceToken'
app_label = 'msync'
verbose_name = '服务令牌'
verbose_name_plural = '服务令牌'
ordering = ['-CreateTime']
def IsExpired(self):
return _utcnow() > self.ExpiresAt.replace(tzinfo=None)
def IsRefreshExpired(self):
return _utcnow() > self.RefreshExpiresAt.replace(tzinfo=None)
def IsValid(self):
return self.TokenStatus == 1 and not self.IsExpired()
def Revoke(self):
self.TokenStatus = 0
self.save(update_fields=['TokenStatus'])
class ObjectGlobalMapping(QModel):
"""对象全局映射表——跨服务对象引用注册"""
OGMUUID = UUIDField(primary_key=True, verbose_name='OGM UUID')
ObjectType = models.CharField(max_length=64, verbose_name='对象类型')
ObjectLocalID = models.CharField(max_length=128, verbose_name='本地ID')
ObjectGlobalID = models.CharField(max_length=128, unique=True, db_index=True, verbose_name='全局ID')
OwnerServiceUUID = models.ForeignKey(
ServiceNode, on_delete=models.CASCADE,
related_name='owned_objects', verbose_name='所属服务UUID',
)
ObjectMetadata = models.JSONField(default=dict, verbose_name='对象元数据')
ObjectStatus = models.SmallIntegerField(default=1, verbose_name='对象状态')
CreateTime = models.DateTimeField(default=_utcnow, verbose_name='创建时间')
UpdateTime = models.DateTimeField(null=True, blank=True, verbose_name='更新时间')
class Meta:
db_table = 'OGM'
app_label = 'msync'
verbose_name = '对象全局映射'
verbose_name_plural = '对象全局映射'
unique_together = [('ObjectType', 'ObjectLocalID', 'OwnerServiceUUID')]
ordering = ['-CreateTime']
indexes = [
models.Index(fields=['ObjectType'], name='idx_ogm_type'),
models.Index(fields=['ObjectGlobalID'], name='idx_ogm_gid'),
]
def __str__(self):
return f'{self.ObjectType}:{self.ObjectGlobalID}'
@property
def IsAlive(self):
return self.ObjectStatus == 1
def Touch(self):
self.UpdateTime = _utcnow()
self.save(update_fields=['UpdateTime'])
class DistributedTransaction(QModel):
"""分布式事务表——跨服务事务编排(Saga 模式)"""
TX_STATUS_PENDING = 0
TX_STATUS_RUNNING = 1
TX_STATUS_COMMITTED = 2
TX_STATUS_COMPENSATING = 3
TX_STATUS_COMPENSATED = 4
TX_STATUS_FAILED = 5
TX_STATUS_TIMED_OUT = 6
TX_STATUS_CHOICES = [
(TX_STATUS_PENDING, '待执行'),
(TX_STATUS_RUNNING, '执行中'),
(TX_STATUS_COMMITTED, '已提交'),
(TX_STATUS_COMPENSATING, '补偿中'),
(TX_STATUS_COMPENSATED, '已补偿'),
(TX_STATUS_FAILED, '失败'),
(TX_STATUS_TIMED_OUT, '超时'),
]
TX_TYPE_SAGA = 'saga'
TX_TYPE_TWO_PHASE = '2pc'
TX_TYPE_CHOICES = [
(TX_TYPE_SAGA, 'Saga'),
(TX_TYPE_TWO_PHASE, '两阶段提交'),
]
TransactionUUID = UUIDField(primary_key=True, verbose_name='事务UUID')
TransactionCode = models.CharField(max_length=64, unique=True, verbose_name='事务编号')
TransactionType = models.CharField(max_length=16, choices=TX_TYPE_CHOICES, default=TX_TYPE_SAGA, verbose_name='事务类型')
TransactionName = models.CharField(max_length=128, verbose_name='事务名称')
TransactionDescription = models.CharField(max_length=512, default='', verbose_name='事务描述')
TransactionStatus = models.SmallIntegerField(
default=TX_STATUS_PENDING, choices=TX_STATUS_CHOICES, verbose_name='事务状态',
)
InitiatorServiceUUID = models.ForeignKey(
ServiceNode, on_delete=models.SET_NULL,
null=True, blank=True, related_name='initiated_dtx', verbose_name='发起服务UUID',
)
ReferenceOGMUUID = UUIDField(null=True, blank=True, verbose_name='关联OGM UUID')
TransactionPayload = models.JSONField(default=dict, verbose_name='事务载荷')
TransactionResult = models.JSONField(default=dict, verbose_name='事务结果')
TimeoutSeconds = models.IntegerField(default=300, verbose_name='超时秒数')
CreateTime = models.DateTimeField(default=_utcnow, verbose_name='创建时间')
UpdateTime = models.DateTimeField(null=True, blank=True, verbose_name='更新时间')
EndTime = models.DateTimeField(null=True, blank=True, verbose_name='结束时间')
class Meta:
db_table = 'DistributedTransaction'
app_label = 'msync'
verbose_name = '分布式事务'
verbose_name_plural = '分布式事务'
ordering = ['-CreateTime']
indexes = [
models.Index(fields=['TransactionStatus'], name='idx_dtx_status'),
models.Index(fields=['InitiatorServiceUUID'], name='idx_dtx_initiator'),
]
def __str__(self):
return self.TransactionCode
@property
def IsTerminal(self):
return self.TransactionStatus in (
self.TX_STATUS_COMMITTED,
self.TX_STATUS_COMPENSATED,
self.TX_STATUS_FAILED,
self.TX_STATUS_TIMED_OUT,
)
@property
def IsTimedOut(self):
if self.TransactionStatus in (self.TX_STATUS_COMMITTED, self.TX_STATUS_COMPENSATED):
return False
if not self.TimeoutSeconds or not self.CreateTime:
return False
elapsed = (_utcnow() - self.CreateTime.replace(tzinfo=None)).total_seconds()
return elapsed > self.TimeoutSeconds
class TransactionStep(QModel):
"""事务步骤表——Saga 编排中的每一步"""
STEP_STATUS_PENDING = 0
STEP_STATUS_RUNNING = 1
STEP_STATUS_COMPLETED = 2
STEP_STATUS_COMPENSATING = 3
STEP_STATUS_COMPENSATED = 4
STEP_STATUS_FAILED = 5
STEP_STATUS_SKIPPED = 6
STEP_STATUS_CHOICES = [
(STEP_STATUS_PENDING, '待执行'),
(STEP_STATUS_RUNNING, '执行中'),
(STEP_STATUS_COMPLETED, '已完成'),
(STEP_STATUS_COMPENSATING, '补偿中'),
(STEP_STATUS_COMPENSATED, '已补偿'),
(STEP_STATUS_FAILED, '失败'),
(STEP_STATUS_SKIPPED, '已跳过'),
]
StepUUID = UUIDField(primary_key=True, verbose_name='步骤UUID')
TransactionUUID = models.ForeignKey(
DistributedTransaction, on_delete=models.CASCADE,
related_name='steps', verbose_name='事务UUID',
)
StepIndex = models.IntegerField(verbose_name='步骤序号')
StepName = models.CharField(max_length=128, verbose_name='步骤名称')
StepDescription = models.CharField(max_length=512, default='', verbose_name='步骤描述')
ParticipantServiceUUID = models.ForeignKey(
ServiceNode, on_delete=models.SET_NULL,
null=True, blank=True, related_name='participated_steps', verbose_name='参与服务UUID',
)
StepStatus = models.SmallIntegerField(
default=STEP_STATUS_PENDING, choices=STEP_STATUS_CHOICES, verbose_name='步骤状态',
)
ActionEndpoint = models.CharField(max_length=512, verbose_name='执行端点')
ActionMethod = models.CharField(max_length=8, default='POST', verbose_name='执行方法')
ActionPayload = models.JSONField(default=dict, verbose_name='执行载荷')
CompensateEndpoint = models.CharField(max_length=512, default='', verbose_name='补偿端点')
CompensateMethod = models.CharField(max_length=8, default='POST', verbose_name='补偿方法')
CompensatePayload = models.JSONField(default=dict, verbose_name='补偿载荷')
StepResult = models.JSONField(default=dict, verbose_name='步骤结果')
RetryCount = models.IntegerField(default=0, verbose_name='重试次数')
MaxRetries = models.IntegerField(default=3, verbose_name='最大重试')
CreateTime = models.DateTimeField(default=_utcnow, verbose_name='创建时间')
StartTime = models.DateTimeField(null=True, blank=True, verbose_name='开始时间')
EndTime = models.DateTimeField(null=True, blank=True, verbose_name='结束时间')
class Meta:
db_table = 'TransactionStep'
app_label = 'msync'
verbose_name = '事务步骤'
verbose_name_plural = '事务步骤'
ordering = ['TransactionUUID', 'StepIndex']
indexes = [
models.Index(fields=['TransactionUUID', 'StepIndex'], name='idx_step_tx_order'),
]
def __str__(self):
return f'{self.StepIndex}: {self.StepName}'
@property
def CanRetry(self):
return self.RetryCount < self.MaxRetries
@property
def IsTerminal(self):
return self.StepStatus in (
self.STEP_STATUS_COMPLETED,
self.STEP_STATUS_COMPENSATED,
self.STEP_STATUS_FAILED,
self.STEP_STATUS_SKIPPED,
)
+263 -13
View File
@@ -1,17 +1,267 @@
"""MSYNC OGM(对象全局映射)模块
"""模块:对象全局映射(OGM)管理
提供跨服务对象引用的注册、解析、更新、注销操作。
OGM 是微服务间共享对象身份的核心机制:
- 每个本地对象注册后获得全局唯一 ID
- 其他服务通过全局 ID 解析到对象归属与本地 ID
- 支持按类型、按服务查询对象映射"""
薄封装:直接转发到 gvsds.apps.msync.ogm 原生层。
"""
import uuid
import datetime
import logging
from gvsdsdk.msync.models import ObjectGlobalMapping, ServiceNode, _utcnow
try:
from gvsds.apps.msync.ogm import OGMManager, OGMError
except ImportError:
raise ImportError(
'OGMManager 仅在主服务器环境可用。'
'子服务器请通过 ServiceClient.ResolveOGM() 使用 HTTP 转发。'
logger = logging.getLogger(__name__)
class OGMError(Exception):
pass
class OGMManager:
"""OGM 管理器——对象全局映射的增删改查"""
@staticmethod
def Register(object_type, local_id, service_name, global_id=None, metadata=None):
"""注册对象全局映射
Args:
object_type: 对象类型(如 'user', 'company', 'asset')
local_id: 对象在所属服务中的本地 ID
service_name: 所属服务名称
global_id: 自定义全局 ID(不指定则自动生成)
metadata: 对象元数据
Returns:
ObjectGlobalMapping 实例
Raises:
OGMError: 注册失败
"""
try:
service = ServiceNode.objects.get(ServiceName=service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise OGMError(f'服务不存在或已禁用: {service_name}')
if ObjectGlobalMapping.objects.filter(
ObjectType=object_type,
ObjectLocalID=local_id,
OwnerServiceUUID=service,
).exists():
raise OGMError(f'对象映射已存在: {object_type}:{local_id}@{service_name}')
if global_id is None:
global_id = f'{object_type}:{uuid.uuid4().hex[:16]}'
if ObjectGlobalMapping.objects.filter(ObjectGlobalID=global_id).exists():
raise OGMError(f'全局 ID 已被占用: {global_id}')
ogm = ObjectGlobalMapping(
OGMUUID=uuid.uuid4().bytes,
ObjectType=object_type,
ObjectLocalID=str(local_id),
ObjectGlobalID=global_id,
OwnerServiceUUID=service,
ObjectMetadata=metadata or {},
ObjectStatus=1,
)
ogm.save()
logger.info(f'OGM 注册: {global_id} -> {object_type}:{local_id}@{service_name}')
return ogm
__all__ = [
'OGMManager',
'OGMError',
]
@staticmethod
def RegisterBatch(items, service_name):
"""批量注册对象全局映射
Args:
items: 列表,每项为 dict,含 ObjectType, LocalID, GlobalID(可选), metadata(可选)
service_name: 所属服务名称
Returns:
注册成功的 ObjectGlobalMapping 列表
"""
try:
service = ServiceNode.objects.get(ServiceName=service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise OGMError(f'服务不存在或已禁用: {service_name}')
results = []
for item in items:
object_type = item['ObjectType']
local_id = str(item['ObjectLocalID'])
global_id = item.get('ObjectGlobalID') or f'{object_type}:{uuid.uuid4().hex[:16]}'
metadata = item.get('Metadata', {})
if ObjectGlobalMapping.objects.filter(
ObjectType=object_type,
ObjectLocalID=local_id,
OwnerServiceUUID=service,
).exists():
logger.warning(f'OGM 跳过(已存在): {object_type}:{local_id}@{service_name}')
continue
ogm = ObjectGlobalMapping(
OGMUUID=uuid.uuid4().bytes,
ObjectType=object_type,
ObjectLocalID=local_id,
ObjectGlobalID=global_id,
OwnerServiceUUID=service,
ObjectMetadata=metadata,
ObjectStatus=1,
)
ogm.save()
results.append(ogm)
logger.info(f'OGM 批量注册: {len(results)}/{len(items)} 成功')
return results
@staticmethod
def Resolve(global_id):
"""通过全局 ID 解析对象映射
Args:
global_id: 全局 ID
Returns:
对象信息字典或 None
"""
try:
ogm = ObjectGlobalMapping.objects.select_related('OwnerServiceUUID').get(
ObjectGlobalID=global_id,
ObjectStatus=1,
)
except ObjectGlobalMapping.DoesNotExist:
return None
return OGMManager._to_dict(ogm)
@staticmethod
def ResolveLocal(object_type, local_id, service_name):
"""通过本地信息解析对象映射
Args:
object_type: 对象类型
local_id: 本地 ID
service_name: 服务名称
Returns:
对象信息字典或 None
"""
try:
service = ServiceNode.objects.get(ServiceName=service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
return None
try:
ogm = ObjectGlobalMapping.objects.select_related('OwnerServiceUUID').get(
ObjectType=object_type,
ObjectLocalID=str(local_id),
OwnerServiceUUID=service,
ObjectStatus=1,
)
except ObjectGlobalMapping.DoesNotExist:
return None
return OGMManager._to_dict(ogm)
@staticmethod
def QueryByType(object_type, active_only=True):
"""按对象类型查询映射
Args:
object_type: 对象类型
active_only: 是否只返回活跃对象
Returns:
对象信息字典列表
"""
qs = ObjectGlobalMapping.objects.select_related('OwnerServiceUUID').filter(ObjectType=object_type)
if active_only:
qs = qs.filter(ObjectStatus=1)
return [OGMManager._to_dict(ogm) for ogm in qs]
@staticmethod
def QueryByService(service_name, object_type=None, active_only=True):
"""按服务查询映射
Args:
service_name: 服务名称
object_type: 可选的对象类型过滤
active_only: 是否只返回活跃对象
Returns:
对象信息字典列表
"""
try:
service = ServiceNode.objects.get(ServiceName=service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
return []
qs = ObjectGlobalMapping.objects.select_related('OwnerServiceUUID').filter(OwnerServiceUUID=service)
if object_type:
qs = qs.filter(ObjectType=object_type)
if active_only:
qs = qs.filter(ObjectStatus=1)
return [OGMManager._to_dict(ogm) for ogm in qs]
@staticmethod
def Update(global_id, metadata=None):
"""更新对象映射的元数据
Args:
global_id: 全局 ID
metadata: 新的元数据(合并更新)
Returns:
更新后的 ObjectGlobalMapping 实例或 None
"""
try:
ogm = ObjectGlobalMapping.objects.get(ObjectGlobalID=global_id, ObjectStatus=1)
except ObjectGlobalMapping.DoesNotExist:
return None
if metadata:
existing = ogm.ObjectMetadata or {}
existing.update(metadata)
ogm.ObjectMetadata = existing
ogm.UpdateTime = _utcnow()
ogm.save()
return ogm
@staticmethod
def Deregister(global_id):
"""注销对象全局映射(软删除)
Args:
global_id: 全局 ID
Returns:
True 如果成功
"""
try:
ogm = ObjectGlobalMapping.objects.get(ObjectGlobalID=global_id, ObjectStatus=1)
except ObjectGlobalMapping.DoesNotExist:
return False
ogm.ObjectStatus = 0
ogm.UpdateTime = _utcnow()
ogm.save(update_fields=['ObjectStatus', 'UpdateTime'])
logger.info(f'OGM 注销: {global_id}')
return True
@staticmethod
def _to_dict(ogm):
"""将 OGM 实例转换为信息字典"""
service = ogm.OwnerServiceUUID
return {
'OGMUUID': str(ogm.OGMUUID) if isinstance(ogm.OGMUUID, bytes) else ogm.OGMUUID,
'ObjectType': ogm.ObjectType,
'ObjectLocalID': ogm.ObjectLocalID,
'ObjectGlobalID': ogm.ObjectGlobalID,
'OwnerService': service.ServiceName if service else None,
'OwnerDomain': service.ServiceDomain if service else None,
'Metadata': ogm.ObjectMetadata,
'Status': ogm.ObjectStatus,
'CreateTime': ogm.CreateTime.isoformat() if ogm.CreateTime else None,
'UpdateTime': ogm.UpdateTime.isoformat() if ogm.UpdateTime else None,
}
+259 -13
View File
@@ -1,17 +1,263 @@
"""MSYNC 注册中心模块
"""模块:服务注册与发现
提供服务节点的注册、注销、发现、心跳等操作。"""
薄封装:直接转发到 gvsds.apps.msync.registry 原生层。
"""
import uuid
import datetime
import logging
from gvsdsdk.msync.models import ServiceNode, _utcnow
try:
from gvsds.apps.msync.registry import ServiceRegistry, RegistryError
except ImportError:
raise ImportError(
'ServiceRegistry 仅在主服务器环境可用。'
'子服务器请通过 ServiceClient.DiscoverService() 使用 HTTP 转发。'
logger = logging.getLogger(__name__)
class RegistryError(Exception):
pass
class ServiceRegistry:
"""服务注册中心——管理微服务的注册、发现与生命周期"""
SERVICE_TYPE_WORKER = 0
SERVICE_TYPE_MASTER = 1
DEFAULT_SCOPE = ['ogm:read', 'ogm:write', 'registry:read']
@staticmethod
def Register(name, domain, secret_hash, raw_secret=None, scope=None, service_type=0, metadata=None):
"""注册新的服务节点(已存在则验证密钥哈希并更新域信息)
Args:
name: 服务名称(唯一标识)
domain: 服务域名或 URL
secret_hash: 服务密钥的 SHA256 哈希(客户端计算,不再传输明文)
raw_secret: 服务原始密钥(用于加密存储,认证时需要)
scope: 权限范围列表
service_type: 服务类型(0=工作节点, 1=主节点)
metadata: 服务元数据
Returns:
ServiceNode 实例
Raises:
RegistryError: 注册失败
"""
import hashlib
# 密钥强制:raw_secret 与 secret_hash 必须自洽;
# 新服务节点必须携带明文密钥,否则后续 HMAC 认证无法完成
if raw_secret:
computed_hash = hashlib.sha256(raw_secret.encode('utf-8')).hexdigest()
if secret_hash and computed_hash != secret_hash:
raise RegistryError('服务密钥哈希不一致')
secret_hash = secret_hash or computed_hash
elif not secret_hash:
raise RegistryError('服务密钥缺失')
existing = ServiceNode.objects.filter(ServiceName=name).first()
if existing:
if not existing.ServiceSecretHash or existing.ServiceSecretHash != secret_hash:
raise RegistryError('服务密钥验证失败')
existing.ServiceDomain = domain
existing.ServiceScope = scope or ServiceRegistry.DEFAULT_SCOPE
existing.ServiceStatus = 1
existing.ServiceType = service_type
existing.ServiceMetadata = metadata or {}
if raw_secret:
existing.SetSecret(raw_secret)
existing.save()
logger.info(f'服务重新注册: {name} ({domain})')
return existing
if not raw_secret:
raise RegistryError('新服务注册必须提供 ServiceSecret 密钥')
node = ServiceNode(
ServiceUUID=uuid.uuid4().bytes,
ServiceName=name,
ServiceDomain=domain,
ServiceScope=scope or ServiceRegistry.DEFAULT_SCOPE,
ServiceStatus=1,
ServiceType=service_type,
ServiceMetadata=metadata or {},
ServiceSecretHash=secret_hash,
)
if raw_secret:
node.SetSecret(raw_secret)
node.save()
logger.info(f'服务注册成功: {name} ({domain})')
return node
__all__ = [
'ServiceRegistry',
'RegistryError',
]
@staticmethod
def Deregister(name):
"""注销服务节点(软删除,设置状态为禁用)
Args:
name: 服务名称
Returns:
True 如果成功
Raises:
RegistryError: 服务不存在
"""
try:
node = ServiceNode.objects.get(ServiceName=name)
except ServiceNode.DoesNotExist:
raise RegistryError(f'服务不存在: {name}')
node.ServiceStatus = 0
node.save(update_fields=['ServiceStatus'])
logger.info(f'服务已注销: {name}')
return True
@staticmethod
def Discover(name):
"""发现指定服务
Args:
name: 服务名称
Returns:
服务信息字典或 None
"""
try:
node = ServiceNode.objects.get(ServiceName=name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
return None
return {
'ServiceName': node.ServiceName,
'ServiceDomain': node.ServiceDomain,
'ServiceScope': node.ServiceScope,
'ServiceType': node.ServiceType,
'ServiceMetadata': node.ServiceMetadata,
'ServiceStatus': node.ServiceStatus,
'LastHeartbeat': node.LastHeartbeat.isoformat() if node.LastHeartbeat else None,
}
@staticmethod
def GetNode(name):
"""获取服务节点实例(不按状态过滤,已注销也可查询);不存在返回 None"""
return ServiceNode.objects.filter(ServiceName=name).first()
@staticmethod
def DiscoverAll(active_only=True, page=1, page_size=50):
"""发现所有已注册服务(分页)
Args:
active_only: 是否只返回活跃服务
page: 页码(从1开始)
page_size: 每页数量
Returns:
包含分页元数据和服务列表的字典
"""
qs = ServiceNode.objects.all()
if active_only:
qs = qs.filter(ServiceStatus=1)
total = qs.count()
offset = (page - 1) * page_size
page_qs = qs[offset:offset + page_size]
services = [
{
'ServiceName': node.ServiceName,
'ServiceDomain': node.ServiceDomain,
'ServiceScope': node.ServiceScope,
'ServiceType': node.ServiceType,
'ServiceMetadata': node.ServiceMetadata,
'ServiceStatus': node.ServiceStatus,
'LastHeartbeat': node.LastHeartbeat.isoformat() if node.LastHeartbeat else None,
}
for node in page_qs
]
return {
'total': total,
'page': page,
'page_size': page_size,
'services': services,
}
@staticmethod
def Heartbeat(name):
"""服务心跳——更新最后心跳时间
Args:
name: 服务名称
Returns:
True 如果成功
Raises:
RegistryError: 服务不存在
"""
try:
node = ServiceNode.objects.get(ServiceName=name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise RegistryError(f'服务不存在或已禁用: {name}')
node.LastHeartbeat = _utcnow()
node.save(update_fields=['LastHeartbeat'])
return True
@staticmethod
def UpdateScope(name, scope):
"""更新服务权限范围
Args:
name: 服务名称
scope: 新的权限范围列表
Raises:
RegistryError: 服务不存在
"""
try:
node = ServiceNode.objects.get(ServiceName=name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise RegistryError(f'服务不存在或已禁用: {name}')
node.ServiceScope = scope
node.save(update_fields=['ServiceScope'])
logger.info(f'服务权限更新: {name} -> {scope}')
return True
@staticmethod
def RotateSecret(name, new_secret):
"""轮换服务密钥
Args:
name: 服务名称
new_secret: 新密钥(明文)
Raises:
RegistryError: 服务不存在
"""
try:
node = ServiceNode.objects.get(ServiceName=name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise RegistryError(f'服务不存在或已禁用: {name}')
node.SetSecret(new_secret)
node.save(update_fields=['ServiceSecretHash', 'ServiceSecretEncrypted'])
from gvsdsdk.msync.models import ServiceToken
ServiceToken.objects.filter(ServiceUUID=node, TokenStatus=1).update(TokenStatus=0)
logger.info(f'服务密钥已轮换: {name}')
return True
@staticmethod
def GetInactiveServices(timeout_minutes=5):
"""获取超时未心跳的服务
Args:
timeout_minutes: 超时分钟数
Returns:
ServiceNode QuerySet
"""
threshold = _utcnow() - datetime.timedelta(minutes=timeout_minutes)
return ServiceNode.objects.filter(
ServiceStatus=1,
LastHeartbeat__lt=threshold,
)
+98
View File
@@ -0,0 +1,98 @@
from rest_framework import serializers
from gvsdsdk.serializer_base import PascalCaseSerializer, PascalCaseModelSerializer
from gvsdsdk.msync.models import ServiceNode
class ServiceNodeSerializer(PascalCaseModelSerializer):
class Meta:
model = ServiceNode
fields = ['ServiceName', 'ServiceDomain', 'ServiceScope', 'ServiceType', 'ServiceMetadata', 'ServiceStatus', 'CreateTime', 'LastHeartbeat']
read_only_fields = ['CreateTime', 'LastHeartbeat']
class ServiceTokenResponseSerializer(PascalCaseSerializer):
AccessToken = serializers.CharField()
RefreshToken = serializers.CharField()
Scope = serializers.ListField()
ExpiresAt = serializers.FloatField()
class AuthTokenRequestSerializer(PascalCaseSerializer):
ServiceName = serializers.CharField()
Payload = serializers.DictField()
Signature = serializers.CharField()
class RefreshTokenRequestSerializer(PascalCaseSerializer):
RefreshToken = serializers.CharField()
class OGMRegisterSerializer(PascalCaseSerializer):
ObjectType = serializers.CharField()
ObjectLocalID = serializers.CharField()
ServiceName = serializers.CharField()
ObjectGlobalID = serializers.CharField(required=False, allow_blank=True)
Metadata = serializers.DictField(required=False, default=dict)
class OGMBatchRegisterSerializer(PascalCaseSerializer):
Items = OGMRegisterSerializer(many=True)
ServiceName = serializers.CharField()
class ServiceRegisterSerializer(PascalCaseSerializer):
ServiceName = serializers.CharField()
ServiceDomain = serializers.CharField()
ServiceSecret = serializers.CharField(write_only=True)
ServiceSecretHash = serializers.CharField(required=False, allow_blank=True)
ServiceScope = serializers.ListField(required=False, default=['ogm:read', 'ogm:write', 'registry:read'])
ServiceType = serializers.IntegerField(default=0)
ServiceMetadata = serializers.DictField(required=False, default=dict)
class TransactionStepSerializer(PascalCaseSerializer):
StepName = serializers.CharField()
StepDescription = serializers.CharField(required=False, default='')
ParticipantServiceUUID = serializers.CharField(required=False, allow_blank=True)
ActionEndpoint = serializers.CharField()
ActionMethod = serializers.CharField(default='POST')
ActionPayload = serializers.DictField(required=False, default=dict)
CompensateEndpoint = serializers.CharField(required=False, default='')
CompensateMethod = serializers.CharField(default='POST')
CompensatePayload = serializers.DictField(required=False, default=dict)
MaxRetries = serializers.IntegerField(default=3)
class TransactionCreateSerializer(PascalCaseSerializer):
TransactionName = serializers.CharField()
TransactionDescription = serializers.CharField(required=False, default='')
TransactionType = serializers.ChoiceField(choices=['saga', '2pc'], default='saga')
Steps = TransactionStepSerializer(many=True)
InitiatorServiceName = serializers.CharField(required=False, allow_blank=True)
ReferenceOGMUUID = serializers.CharField(required=False, allow_blank=True)
Payload = serializers.DictField(required=False, default=dict)
TimeoutSeconds = serializers.IntegerField(default=300)
class TransactionExecuteSerializer(PascalCaseSerializer):
AccessToken = serializers.CharField(required=False, allow_blank=True)
class TransactionRetryStepSerializer(PascalCaseSerializer):
StepIndex = serializers.IntegerField()
AccessToken = serializers.CharField(required=False, allow_blank=True)
class LocalExecuteSerializer(PascalCaseSerializer):
StepName = serializers.CharField()
Payload = serializers.DictField(required=False, default=dict)
TransactionCode = serializers.CharField(required=False, allow_blank=True)
StepIndex = serializers.IntegerField(required=False, allow_null=True)
class LocalCompensateSerializer(PascalCaseSerializer):
StepName = serializers.CharField()
Payload = serializers.DictField(required=False, default=dict)
ActionResult = serializers.DictField(required=False, default=dict)
TransactionCode = serializers.CharField(required=False, allow_blank=True)
StepIndex = serializers.IntegerField(required=False, allow_null=True)
+545
View File
@@ -0,0 +1,545 @@
"""模块:MSYNC 服务端(Socket.IO)
替代 REST API,提供实时双向通信。事件协议字段全部使用 PascalCase。
事件协议:
客户端 → 服务端:
msync:register 注册服务(bootstrap,必须携带密钥)
msync:auth HMAC 签名认证(密钥验证,成功后才可调用业务事件)
msync:auth:refresh 刷新令牌
msync:heartbeat 应用层心跳(需已认证)
msync:deregister 注销服务(需已认证,仅限本服务)
msync:discover 发现所有服务(需已认证)
msync:discover:one 发现单个服务(需已认证)
msync:ogm:register 注册 OGM(需已认证)
msync:ogm:batch 批量注册 OGM(需已认证)
msync:ogm:resolve 解析 OGM(需已认证)
msync:ogm:query 查询 OGM(需已认证)
msync:ogm:update 更新 OGM(需已认证)
msync:ogm:deregister 注销 OGM(需已认证)
msync:tx:create 创建事务(需已认证)
msync:tx:execute 执行事务(需已认证)
msync:tx:compensate 补偿事务(需已认证)
msync:tx:cancel 取消事务(需已认证)
msync:tx:detail 事务详情(需已认证)
msync:tx:retry 重试步骤(需已认证)
msync:tx:retry_compensate 重试补偿(需已认证)
msync:tx:timeout_check 超时检测(需已认证)
msync:tx:query 按状态查询(需已认证)
鉴权规则: 除 register / auth / auth:refresh 外,所有业务事件必须
先通过 msync:auth 密钥认证(HMAC 签名验证),否则返回 401。
服务端 → 客户端(推送):
msync:service:online 服务上线
msync:service:offline 服务下线
"""
import logging
import threading
import traceback
import socketio
logger = logging.getLogger(__name__)
NS = '/msync'
sio = socketio.Server(async_mode='gevent', cors_allowed_origins=[
'https://*.gvsds.com',
'https://*.gvsds.dev',
'https://*.yc8z.org',
'https://*.ycbz.org',
'https://*.gvsds.org',
'https://*.gvsds.studio',
'http://localhost:*',
'http://127.0.0.1:*',
], logger=True, engineio_logger=True)
_sid_service_map = {}
_service_sid_map = {}
_sid_auth_map = {}
_map_lock = threading.Lock()
def _ok(data=None, msg='success'):
return {'code': 200, 'msg': msg, 'data': data}
def _err(msg, code=400):
return {'code': code, 'msg': msg, 'data': None}
def _d(data, key, default=None):
"""安全获取字典字段"""
if not isinstance(data, dict):
return default
return data.get(key, default)
def _RequireAuth(sid):
"""校验 sid 是否已完成密钥认证(msync:auth);通过返回服务名,否则返回 None"""
with _map_lock:
return _sid_auth_map.get(sid)
@sio.on('connect', namespace=NS)
def on_connect(sid, environ):
logger.info(f'[SIO] → connect sid={sid}')
@sio.on('disconnect', namespace=NS)
def on_disconnect(sid):
with _map_lock:
service_name = _sid_service_map.pop(sid, None)
if service_name:
_service_sid_map.pop(service_name, None)
_sid_auth_map.pop(sid, None)
if service_name:
try:
from gvsdsdk.msync.registry import ServiceRegistry
ServiceRegistry.Deregister(service_name)
except Exception:
pass
sio.emit('msync:service:offline', {'Service': service_name}, namespace=NS, skip_sid=sid)
logger.info(f'[SIO] 服务断开: {service_name} ({sid})')
else:
logger.info(f'[SIO] 客户端断开: {sid}')
@sio.on('msync:register', namespace=NS)
def on_register(sid, data):
logger.info(f'[SIO] → msync:register sid={sid} name={_d(data, "ServiceName")}')
from gvsdsdk.msync.registry import ServiceRegistry, RegistryError
try:
node = ServiceRegistry.Register(
name=_d(data, 'ServiceName', ''),
domain=_d(data, 'ServiceDomain', ''),
secret_hash=_d(data, 'ServiceSecretHash', ''),
raw_secret=_d(data, 'ServiceSecret'),
scope=_d(data, 'ServiceScope'),
service_type=_d(data, 'ServiceType', 0),
metadata=_d(data, 'ServiceMetadata', {}),
)
except RegistryError as e:
return _err(str(e), 409)
except Exception as e:
logger.error(f'注册异常: {e}\n{traceback.format_exc()}')
return _err(str(e), 500)
service_name = _d(data, 'ServiceName')
with _map_lock:
_sid_service_map[sid] = service_name
_service_sid_map[service_name] = sid
# 若以不同服务名重新注册,旧认证状态失效
if _sid_auth_map.get(sid) != service_name:
_sid_auth_map.pop(sid, None)
sio.emit('msync:service:online', {
'ServiceName': service_name,
'ServiceDomain': _d(data, 'ServiceDomain'),
}, namespace=NS, skip_sid=sid)
from gvsdsdk.msync.serializers import ServiceNodeSerializer
result = _ok(ServiceNodeSerializer(node).data)
logger.info(f'[SIO] ← msync:register sid={sid} code=200')
return result
@sio.on('msync:auth', namespace=NS)
def on_auth(sid, data):
service_name = _d(data, 'ServiceName')
logger.info(f'[SIO] → msync:auth sid={sid} service={service_name}')
from gvsdsdk.msync.auth import AuthenticateService
try:
token = AuthenticateService(
service_name=service_name,
signed_payload=_d(data, 'Payload'),
signature=_d(data, 'Signature'),
)
if token is None:
return _err('认证失败', 401)
try:
scope = token.TokenScope
if not isinstance(scope, list):
scope = list(scope) if scope else []
except Exception:
scope = []
try:
from django.utils.dateparse import parse_datetime
if isinstance(token.ExpiresAt, str):
dt = parse_datetime(token.ExpiresAt)
if dt is None:
raise ValueError
else:
dt = token.ExpiresAt
# 若是 naive datetime,按 UTC 处理(DB 存的就是 naive UTC)
if dt.tzinfo is None:
import datetime as _dt
dt = dt.replace(tzinfo=_dt.timezone.utc)
expires_ts = dt.timestamp()
except Exception:
import time as _time
expires_ts = _time.time() + 7200
with _map_lock:
_sid_auth_map[sid] = service_name
return _ok({
'AccessToken': token.AccessToken,
'RefreshToken': token.RefreshToken,
'Scope': scope,
'ExpiresAt': expires_ts,
})
except Exception as e:
logger.error(f'[SIO] msync:auth 异常: {e}\n{traceback.format_exc()}')
return _err(str(e), 500)
@sio.on('msync:auth:refresh', namespace=NS)
def on_auth_refresh(sid, data):
from gvsdsdk.msync.auth import RefreshServiceToken
try:
token = RefreshServiceToken(_d(data, 'RefreshToken', ''))
if token is None:
return _err('刷新失败', 401)
try:
scope = token.TokenScope
if not isinstance(scope, list):
scope = list(scope) if scope else []
except Exception:
scope = []
try:
from django.utils.dateparse import parse_datetime
if isinstance(token.ExpiresAt, str):
dt = parse_datetime(token.ExpiresAt)
if dt is None:
raise ValueError
else:
dt = token.ExpiresAt
if dt.tzinfo is None:
import datetime as _dt
dt = dt.replace(tzinfo=_dt.timezone.utc)
expires_ts = dt.timestamp()
except Exception:
import time as _time
expires_ts = _time.time() + 7200
with _map_lock:
_sid_auth_map[sid] = token.ServiceUUID.ServiceName
return _ok({
'AccessToken': token.AccessToken,
'RefreshToken': token.RefreshToken,
'Scope': scope,
'ExpiresAt': expires_ts,
})
except Exception as e:
logger.error(f'Socket令牌刷新异常: {e}\n{traceback.format_exc()}')
return _err(str(e), 500)
@sio.on('msync:heartbeat', namespace=NS)
def on_heartbeat(sid, data):
with _map_lock:
service_name = _sid_service_map.get(sid)
auth_name = _sid_auth_map.get(sid)
if not service_name:
return _err('未注册', 401)
if not auth_name or auth_name != service_name:
return _err('需要密钥认证', 401)
from gvsdsdk.msync.registry import ServiceRegistry, RegistryError
try:
ServiceRegistry.Heartbeat(service_name)
except RegistryError as e:
return _err(str(e), 404)
return _ok(msg='心跳成功')
@sio.on('msync:deregister', namespace=NS)
def on_deregister(sid, data):
with _map_lock:
service_name = _sid_service_map.pop(sid, None)
if service_name:
_service_sid_map.pop(service_name, None)
auth_name = _sid_auth_map.pop(sid, None)
name = _d(data, 'ServiceName') or service_name
if not name:
return _err('未指定服务名', 400)
# 仅允许注销本服务(防止伪造 ServiceName 注销他人)
if not auth_name or auth_name != name:
return _err('需要密钥认证', 401)
from gvsdsdk.msync.registry import ServiceRegistry, RegistryError
try:
ServiceRegistry.Deregister(name)
except RegistryError as e:
return _err(str(e), 404)
sio.emit('msync:service:offline', {'Service': name}, namespace=NS, skip_sid=sid)
return _ok(msg='注销成功')
@sio.on('msync:discover', namespace=NS)
def on_discover(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.registry import ServiceRegistry
active_only = _d(data, 'ActiveOnly', True)
if not isinstance(active_only, bool):
active_only = True
services = ServiceRegistry.DiscoverAll(active_only=active_only)
return _ok(services)
@sio.on('msync:discover:one', namespace=NS)
def on_discover_one(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.registry import ServiceRegistry
info = ServiceRegistry.Discover(_d(data, 'ServiceName', ''))
if info is None:
return _err('服务不存在', 404)
return _ok(info)
@sio.on('msync:ogm:register', namespace=NS)
def on_ogm_register(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.ogm import OGMManager, OGMError
try:
ogm = OGMManager.Register(
object_type=_d(data, 'ObjectType'),
local_id=_d(data, 'ObjectLocalID'),
service_name=_d(data, 'ServiceName'),
global_id=_d(data, 'ObjectGlobalID'),
metadata=_d(data, 'Metadata', {}),
)
except OGMError as e:
return _err(str(e), 409)
except Exception as e:
logger.error(f'OGM注册异常: {e}\n{traceback.format_exc()}')
return _err(str(e), 500)
return _ok(OGMManager._to_dict(ogm))
@sio.on('msync:ogm:batch', namespace=NS)
def on_ogm_batch(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.ogm import OGMManager, OGMError
try:
results = OGMManager.RegisterBatch(
items=_d(data, 'Items', []),
service_name=_d(data, 'ServiceName'),
)
except OGMError as e:
return _err(str(e), 409)
return _ok([OGMManager._to_dict(o) for o in results])
@sio.on('msync:ogm:resolve', namespace=NS)
def on_ogm_resolve(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.ogm import OGMManager
info = OGMManager.Resolve(_d(data, 'ObjectGlobalID', ''))
if info is None:
return _err('映射不存在', 404)
return _ok(info)
@sio.on('msync:ogm:query', namespace=NS)
def on_ogm_query(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.ogm import OGMManager
object_type = _d(data, 'ObjectType')
service_name = _d(data, 'ServiceName')
local_id = _d(data, 'ObjectLocalID')
if local_id and object_type and service_name:
info = OGMManager.ResolveLocal(object_type, local_id, service_name)
if info is None:
return _err('映射不存在', 404)
return _ok(info)
if object_type:
results = OGMManager.QueryByType(object_type)
return _ok(results)
if service_name:
results = OGMManager.QueryByService(service_name, object_type=object_type)
return _ok(results)
return _err('请提供 ObjectType 或 ServiceName 参数', 400)
@sio.on('msync:ogm:update', namespace=NS)
def on_ogm_update(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.ogm import OGMManager
ogm = OGMManager.Update(_d(data, 'ObjectGlobalID', ''), metadata=_d(data, 'Metadata'))
if ogm is None:
return _err('映射不存在', 404)
return _ok(OGMManager._to_dict(ogm))
@sio.on('msync:ogm:deregister', namespace=NS)
def on_ogm_deregister(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.ogm import OGMManager
success = OGMManager.Deregister(_d(data, 'ObjectGlobalID', ''))
if not success:
return _err('映射不存在', 404)
return _ok(msg='注销成功')
@sio.on('msync:tx:create', namespace=NS)
def on_tx_create(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
try:
dtx = TransactionManager.Create(
name=_d(data, 'TransactionName'),
steps_data=_d(data, 'Steps'),
description=_d(data, 'TransactionDescription', ''),
transaction_type=_d(data, 'TransactionType', 'saga'),
initiator_service_name=_d(data, 'InitiatorServiceName'),
reference_ogm_uuid=_d(data, 'ReferenceOGMUUID'),
payload=_d(data, 'Payload'),
timeout_seconds=_d(data, 'TimeoutSeconds', 300),
)
except TransactionError as e:
return _err(str(e), 400)
return _ok(TransactionManager.GetDetail(dtx.TransactionCode))
@sio.on('msync:tx:execute', namespace=NS)
def on_tx_execute(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
try:
dtx = TransactionManager.Execute(
_d(data, 'TransactionCode', ''),
access_token=_d(data, 'AccessToken'),
)
except TransactionError as e:
return _err(str(e), 400)
return _ok(TransactionManager.GetDetail(dtx.TransactionCode))
@sio.on('msync:tx:compensate', namespace=NS)
def on_tx_compensate(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
try:
dtx = TransactionManager.Compensate(
_d(data, 'TransactionCode', ''),
access_token=_d(data, 'AccessToken'),
)
except TransactionError as e:
return _err(str(e), 400)
return _ok(TransactionManager.GetDetail(dtx.TransactionCode))
@sio.on('msync:tx:cancel', namespace=NS)
def on_tx_cancel(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
try:
TransactionManager.Cancel(_d(data, 'TransactionCode', ''))
except TransactionError as e:
return _err(str(e), 400)
return _ok(msg='事务已取消')
@sio.on('msync:tx:detail', namespace=NS)
def on_tx_detail(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager
detail = TransactionManager.GetDetail(_d(data, 'TransactionCode', ''))
if detail is None:
return _err('事务不存在', 404)
return _ok(detail)
@sio.on('msync:tx:retry', namespace=NS)
def on_tx_retry(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
try:
step = TransactionManager.RetryStep(
_d(data, 'TransactionCode', ''),
step_index=_d(data, 'StepIndex', 0),
access_token=_d(data, 'AccessToken'),
)
except TransactionError as e:
return _err(str(e), 400)
return _ok({
'StepIndex': step.StepIndex,
'StepName': step.StepName,
'Status': step.get_StepStatus_display(),
'RetryCount': step.RetryCount,
})
@sio.on('msync:tx:retry_compensate', namespace=NS)
def on_tx_retry_compensate(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
try:
step = TransactionManager.RetryCompensate(
_d(data, 'TransactionCode', ''),
step_index=_d(data, 'StepIndex', 0),
access_token=_d(data, 'AccessToken'),
)
except TransactionError as e:
return _err(str(e), 400)
return _ok({
'StepIndex': step.StepIndex,
'StepName': step.StepName,
'Status': step.get_StepStatus_display(),
'RetryCount': step.RetryCount,
'TransactionStatus': step.TransactionUUID.get_TransactionStatus_display(),
})
@sio.on('msync:tx:timeout_check', namespace=NS)
def on_tx_timeout_check(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager
timed_out = TransactionManager.CheckTimeouts()
return _ok({'TimedOut': timed_out, 'Count': len(timed_out)})
@sio.on('msync:tx:query', namespace=NS)
def on_tx_query(sid, data):
if not _RequireAuth(sid):
return _err('需要密钥认证', 401)
from gvsdsdk.msync.transaction import TransactionManager
status = _d(data, 'Status')
if status is None:
return _err('请提供 Status 参数', 400)
try:
status_int = int(status)
except (ValueError, TypeError):
return _err('Status 必须为整数', 400)
results = TransactionManager.QueryByStatus(status_int)
return _ok(results)
def create_sio_app(django_app):
return socketio.WSGIApp(sio, django_app)
+664 -14
View File
@@ -1,18 +1,668 @@
"""MSYNC 分布式事务模块
"""模块:远程事务管理器
基于 Saga 模式的跨服务事务编排引擎,运行于主服务器(账本服务器)。
通过调用各工作服务的本地事务管理器端点来执行与补偿步骤。
薄封装:直接转发到 gvsds.apps.msync.transaction 原生层。
"""
架构:
主服务器 (TransactionManager)
│
├──▶ POST {service_domain}/api/msync/local/execute/
│ {StepName, payload}
│ → 工作服务的 LocalTransactionManager.execute_step()
│
└──▶ POST {service_domain}/api/msync/local/compensate/
{StepName, payload, ActionResult}
→ 工作服务的 LocalTransactionManager.compensate_step()
try:
from gvsds.apps.msync.transaction import TransactionManager, TransactionError, StepExecutionError
except ImportError:
raise ImportError(
'TransactionManager 仅在主服务器环境可用。'
'子服务器请通过 LocalTransactionManager 注册本地步骤处理器。'
事务生命周期:
Pending → Running → Committed
↘ Failed → Compensating → Compensated
↘ TimedOut → Compensating → Compensated"""
import uuid
import json
import logging
import os
import datetime
import requests
from requests.adapters import HTTPAdapter
from urllib3.util.retry import Retry
from django.db import transaction as db_transaction
from gvsdsdk.msync.models import (
DistributedTransaction,
TransactionStep,
ServiceNode,
_utcnow,
)
logger = logging.getLogger(__name__)
LOCAL_EXECUTE_PATH = '/api/msync/local/execute/'
LOCAL_COMPENSATE_PATH = '/api/msync/local/compensate/'
# ─── 模块级 HTTP 会话(连接池 + 重试 + TLS 验证)─────────────────
_http_session = requests.Session()
_http_session.verify = True # 显式启用 TLS 证书验证
# 支持通过环境变量指定 CA 证书包(开发/部署灵活性)
_ca_bundle = os.environ.get('REQUESTS_CA_BUNDLE') or os.environ.get('CURL_CA_BUNDLE')
if _ca_bundle:
_http_session.verify = _ca_bundle
# 配置连接池重试策略
_retry_adapter = HTTPAdapter(
max_retries=Retry(
total=3,
backoff_factor=0.5,
status_forcelist=[502, 503, 504],
allowed_methods=['POST'],
),
pool_connections=10,
pool_maxsize=10,
)
_http_session.mount('https://', _retry_adapter)
_http_session.mount('http://', _retry_adapter)
class TransactionError(Exception):
pass
class StepExecutionError(Exception):
def __init__(self, message, step_index=None, status_code=None, response_body=None):
super().__init__(message)
self.StepIndex = step_index
self.StatusCode = status_code
self.ResponseBody = response_body
class TransactionManager:
"""远程事务管理器——Saga 编排引擎(运行于主服务器)
调用各工作服务的本地事务管理器来执行步骤:
- 执行步骤 → POST {service_domain}/api/msync/local/execute/
- 补偿步骤 → POST {service_domain}/api/msync/local/compensate/
"""
@staticmethod
@db_transaction.atomic
def Create(name, steps_data, description='', transaction_type='saga',
initiator_service_name=None, reference_ogm_uuid=None,
payload=None, timeout_seconds=300):
"""创建分布式事务
Args:
name: 事务名称
steps_data: 步骤数据列表,每项为 dict:
{
'name': '步骤名称(对应本地事务管理器注册的处理器名)',
'description': '步骤描述',
'ParticipantService': '服务名称(账本中注册的服务名)',
'ActionPayload': {...},
'CompensatePayload': {...},
'MaxRetries': 3,
}
description: 事务描述
transaction_type: 事务类型 (saga/2pc)
initiator_service_name: 发起服务名称
reference_ogm_uuid: 关联 OGM UUID
payload: 事务级载荷
timeout_seconds: 超时秒数
Returns:
DistributedTransaction 实例
"""
if not steps_data:
raise TransactionError('事务至少需要一个步骤')
initiator = None
if initiator_service_name:
try:
initiator = ServiceNode.objects.get(ServiceName=initiator_service_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise TransactionError(f'发起服务不存在: {initiator_service_name}')
tx_code = f'DTX-{uuid.uuid4().hex[:12].upper()}'
dtx = DistributedTransaction(
TransactionUUID=uuid.uuid4().bytes,
TransactionCode=tx_code,
TransactionType=transaction_type,
TransactionName=name,
TransactionDescription=description,
TransactionStatus=DistributedTransaction.TX_STATUS_PENDING,
InitiatorServiceUUID=initiator,
ReferenceOGMUUID=reference_ogm_uuid,
TransactionPayload=payload or {},
TimeoutSeconds=timeout_seconds,
)
dtx.save()
for idx, step_data in enumerate(steps_data):
participant = None
participant_name = step_data.get('ParticipantService')
if participant_name:
try:
participant = ServiceNode.objects.get(ServiceName=participant_name, ServiceStatus=1)
except ServiceNode.DoesNotExist:
raise TransactionError(f'参与服务不存在: {participant_name}')
step = TransactionStep(
StepUUID=uuid.uuid4().bytes,
TransactionUUID=dtx,
StepIndex=idx,
StepName=step_data['name'],
StepDescription=step_data.get('description', ''),
ParticipantServiceUUID=participant,
StepStatus=TransactionStep.STEP_STATUS_PENDING,
ActionEndpoint=step_data.get('ActionEndpoint', ''),
ActionMethod='POST',
ActionPayload=step_data.get('ActionPayload', {}),
CompensateEndpoint=step_data.get('CompensateEndpoint', ''),
CompensateMethod='POST',
CompensatePayload=step_data.get('CompensatePayload', {}),
MaxRetries=step_data.get('MaxRetries', 3),
)
step.save()
logger.info(f'分布式事务创建: {tx_code} ({name}), {len(steps_data)} 步骤')
return dtx
@staticmethod
@db_transaction.atomic
def Execute(transaction_code, access_token=None):
"""执行分布式事务(Saga 正向流程)
按步骤顺序依次调用各工作服务的本地事务管理器,
任一步骤失败则触发补偿流程。
Args:
transaction_code: 事务代码
access_token: 服务间认证令牌
Returns:
DistributedTransaction 实例
"""
try:
dtx = DistributedTransaction.objects.select_for_update().get(TransactionCode=transaction_code)
except DistributedTransaction.DoesNotExist:
raise TransactionError(f'事务不存在: {transaction_code}')
if dtx.TransactionStatus not in (
DistributedTransaction.TX_STATUS_PENDING,
DistributedTransaction.TX_STATUS_RUNNING,
):
raise TransactionError(f'事务状态不允许执行: {dtx.get_TransactionStatus_display()}')
if dtx.IsTimedOut:
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_TIMED_OUT
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
db_transaction.on_commit(
lambda: TransactionManager.Compensate(transaction_code, access_token=access_token)
)
raise TransactionError(f'事务已超时: {transaction_code}')
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_RUNNING
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
steps = list(dtx.steps.all().order_by('StepIndex'))
for step in steps:
if step.StepStatus == TransactionStep.STEP_STATUS_COMPLETED:
continue
if dtx.IsTimedOut:
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_TIMED_OUT
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
db_transaction.on_commit(
lambda tc=transaction_code, at=access_token: TransactionManager.Compensate(tc, access_token=at)
)
raise TransactionError(f'事务执行超时: {transaction_code}')
try:
result = TransactionManager._execute_step(step, access_token=access_token)
step.StepStatus = TransactionStep.STEP_STATUS_COMPLETED
step.StepResult = result
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
except StepExecutionError as e:
step.StepStatus = TransactionStep.STEP_STATUS_FAILED
step.StepResult = {
'error': str(e),
'status_code': e.StatusCode,
'response': e.ResponseBody,
}
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
logger.warning(f'事务步骤失败: {transaction_code} step={step.StepIndex}, 触发补偿')
db_transaction.on_commit(
lambda tc=transaction_code, at=access_token: TransactionManager.Compensate(tc, access_token=at)
)
return DistributedTransaction.objects.get(TransactionCode=transaction_code)
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_COMMITTED
dtx.EndTime = _utcnow()
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'EndTime', 'UpdateTime'])
logger.info(f'分布式事务提交: {transaction_code}')
return dtx
@staticmethod
@db_transaction.atomic
def Compensate(transaction_code, access_token=None):
"""补偿分布式事务(Saga 反向流程)
逆序遍历已完成步骤,调用各工作服务的本地事务管理器补偿端点。
Args:
transaction_code: 事务代码
access_token: 服务间认证令牌
"""
try:
dtx = DistributedTransaction.objects.select_for_update().get(TransactionCode=transaction_code)
except DistributedTransaction.DoesNotExist:
raise TransactionError(f'事务不存在: {transaction_code}')
if dtx.TransactionStatus == DistributedTransaction.TX_STATUS_COMPENSATED:
return dtx
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_COMPENSATING
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
completed_steps = list(
dtx.steps.filter(StepStatus=TransactionStep.STEP_STATUS_COMPLETED)
.order_by('-StepIndex')
)
__all__ = [
'TransactionManager',
'TransactionError',
'StepExecutionError',
]
compensation_failed = False
for step in completed_steps:
step.StepStatus = TransactionStep.STEP_STATUS_COMPENSATING
step.save(update_fields=['StepStatus'])
try:
result = TransactionManager._compensate_step(step, access_token=access_token)
step.StepStatus = TransactionStep.STEP_STATUS_COMPENSATED
step.StepResult.update({'compensate_result': result})
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
except StepExecutionError as e:
# 补偿失败:步骤置 FAILED(而非停留在 COMPENSATING),
# 使 RetryCompensate 可再次补偿,避免事务永远补不回来
compensation_failed = True
step.StepStatus = TransactionStep.STEP_STATUS_FAILED
step.StepResult.update({
'compensate_error': str(e),
'compensate_status_code': e.StatusCode,
})
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
logger.error(f'补偿步骤失败: {transaction_code} step={step.StepIndex}: {e}')
if compensation_failed:
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_FAILED
else:
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_COMPENSATED
dtx.EndTime = _utcnow()
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'EndTime', 'UpdateTime'])
logger.info(f'分布式事务补偿完成: {transaction_code} -> {dtx.get_TransactionStatus_display()}')
return dtx
@staticmethod
def RetryStep(transaction_code, step_index, access_token=None):
"""重试失败步骤"""
try:
dtx = DistributedTransaction.objects.get(TransactionCode=transaction_code)
except DistributedTransaction.DoesNotExist:
raise TransactionError(f'事务不存在: {transaction_code}')
try:
step = dtx.steps.get(StepIndex=step_index)
except TransactionStep.DoesNotExist:
raise TransactionError(f'步骤不存在: {step_index}')
if step.StepStatus not in (
TransactionStep.STEP_STATUS_FAILED,
TransactionStep.STEP_STATUS_PENDING,
):
raise TransactionError(f'步骤状态不允许重试: {step.get_StepStatus_display()}')
if not step.CanRetry:
raise TransactionError(f'步骤已达到最大重试次数: {step.MaxRetries}')
step.RetryCount += 1
step.StepStatus = TransactionStep.STEP_STATUS_PENDING
step.save(update_fields=['RetryCount', 'StepStatus'])
try:
result = TransactionManager._execute_step(step, access_token=access_token)
step.StepStatus = TransactionStep.STEP_STATUS_COMPLETED
step.StepResult = result
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
except StepExecutionError as e:
step.StepStatus = TransactionStep.STEP_STATUS_FAILED
step.StepResult = {
'error': str(e),
'status_code': e.StatusCode,
'response': e.ResponseBody,
'retry_count': step.RetryCount,
}
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
return step
@staticmethod
@db_transaction.atomic
def RetryCompensate(transaction_code, step_index, access_token=None):
"""重试补偿失败的步骤
仅允许对补偿失败(StepResult 含 compensate_error)的步骤重试补偿;
补偿成功后若不存在仍补偿失败的步骤,事务置 Compensated。
Args:
transaction_code: 事务代码
step_index: 步骤序号
access_token: 服务间认证令牌
Returns:
TransactionStep 实例
"""
try:
dtx = DistributedTransaction.objects.select_for_update().get(TransactionCode=transaction_code)
except DistributedTransaction.DoesNotExist:
raise TransactionError(f'事务不存在: {transaction_code}')
if dtx.TransactionStatus == DistributedTransaction.TX_STATUS_COMPENSATED:
raise TransactionError('事务已补偿完成')
try:
step = dtx.steps.get(StepIndex=step_index)
except TransactionStep.DoesNotExist:
raise TransactionError(f'步骤不存在: {step_index}')
if step.StepStatus != TransactionStep.STEP_STATUS_FAILED:
raise TransactionError(f'仅补偿失败的步骤可重试补偿: {step.get_StepStatus_display()}')
if 'compensate_error' not in (step.StepResult or {}):
raise TransactionError('该步骤为正向执行失败,请使用 RetryStep 重试执行')
if not step.CanRetry:
raise TransactionError(f'步骤已达到最大重试次数: {step.MaxRetries}')
step.RetryCount += 1
step.StepStatus = TransactionStep.STEP_STATUS_COMPENSATING
step.save(update_fields=['RetryCount', 'StepStatus'])
try:
result = TransactionManager._compensate_step(step, access_token=access_token)
step.StepStatus = TransactionStep.STEP_STATUS_COMPENSATED
step.StepResult.update({'compensate_result': result})
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
except StepExecutionError as e:
step.StepStatus = TransactionStep.STEP_STATUS_FAILED
step.StepResult.update({
'compensate_error': str(e),
'compensate_status_code': e.StatusCode,
})
step.EndTime = _utcnow()
step.save(update_fields=['StepStatus', 'StepResult', 'EndTime'])
logger.error(f'补偿重试失败: {transaction_code} step={step.StepIndex}: {e}')
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_FAILED
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
return step
# 补偿成功:若仍存在补偿失败的步骤则事务保持 FAILED,否则 COMPENSATED
remaining_failed = any(
'compensate_error' in (s.StepResult or {})
for s in dtx.steps.all()
)
if remaining_failed:
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_FAILED
else:
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_COMPENSATED
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
logger.info(f'补偿重试完成: {transaction_code} step={step.StepIndex} -> {dtx.get_TransactionStatus_display()}')
return step
@staticmethod
def GetDetail(transaction_code):
"""获取事务详情(含所有步骤)"""
try:
dtx = DistributedTransaction.objects.prefetch_related('steps').get(
TransactionCode=transaction_code,
)
except DistributedTransaction.DoesNotExist:
return None
steps = []
for step in dtx.steps.all().order_by('StepIndex'):
steps.append({
'Index': step.StepIndex,
'Name': step.StepName,
'Description': step.StepDescription,
'Participant': step.ParticipantServiceUUID.ServiceName if step.ParticipantServiceUUID else None,
'Status': step.get_StepStatus_display(),
'StatusCode': step.StepStatus,
'Result': step.StepResult,
'RetryCount': step.RetryCount,
'MaxRetries': step.MaxRetries,
'StartTime': step.StartTime.isoformat() if step.StartTime else None,
'EndTime': step.EndTime.isoformat() if step.EndTime else None,
})
return {
'TransactionCode': dtx.TransactionCode,
'Name': dtx.TransactionName,
'Description': dtx.TransactionDescription,
'Type': dtx.TransactionType,
'Status': dtx.get_TransactionStatus_display(),
'StatusCode': dtx.TransactionStatus,
'Initiator': dtx.InitiatorServiceUUID.ServiceName if dtx.InitiatorServiceUUID else None,
'Payload': dtx.TransactionPayload,
'Result': dtx.TransactionResult,
'TimeoutSeconds': dtx.TimeoutSeconds,
'Steps': steps,
'CreateTime': dtx.CreateTime.isoformat(),
'EndTime': dtx.EndTime.isoformat() if dtx.EndTime else None,
}
@staticmethod
def Cancel(transaction_code):
"""取消待执行的事务"""
try:
dtx = DistributedTransaction.objects.get(TransactionCode=transaction_code)
except DistributedTransaction.DoesNotExist:
raise TransactionError(f'事务不存在: {transaction_code}')
if dtx.TransactionStatus != DistributedTransaction.TX_STATUS_PENDING:
raise TransactionError(f'只有待执行状态的事务可以取消')
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_FAILED
dtx.EndTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'EndTime'])
dtx.steps.filter(StepStatus=TransactionStep.STEP_STATUS_PENDING).update(
StepStatus=TransactionStep.STEP_STATUS_SKIPPED,
EndTime=_utcnow(),
)
logger.info(f'分布式事务取消: {transaction_code}')
return dtx
@staticmethod
def CheckTimeouts():
"""检测所有超时事务并触发补偿"""
timed_out = []
running_txs = DistributedTransaction.objects.filter(
TransactionStatus__in=(
DistributedTransaction.TX_STATUS_PENDING,
DistributedTransaction.TX_STATUS_RUNNING,
),
)
for dtx in running_txs:
if dtx.IsTimedOut:
TransactionManager._mark_timed_out(dtx)
try:
TransactionManager.Compensate(dtx.TransactionCode)
except TransactionError:
pass
timed_out.append(dtx.TransactionCode)
if timed_out:
logger.warning(f'检测到 {len(timed_out)} 个超时事务: {timed_out}')
return timed_out
@staticmethod
def QueryByStatus(status, limit=50):
"""按状态查询事务"""
txs = DistributedTransaction.objects.filter(
TransactionStatus=status,
).order_by('-CreateTime')[:limit]
return [
{
'TransactionCode': tx.TransactionCode,
'Name': tx.TransactionName,
'Type': tx.TransactionType,
'Status': tx.get_TransactionStatus_display(),
'Initiator': tx.InitiatorServiceUUID.ServiceName if tx.InitiatorServiceUUID else None,
'CreateTime': tx.CreateTime.isoformat(),
'EndTime': tx.EndTime.isoformat() if tx.EndTime else None,
}
for tx in txs
]
# ─── 内部方法 ───────────────────────────────────────────
@staticmethod
def _mark_timed_out(dtx):
dtx.TransactionStatus = DistributedTransaction.TX_STATUS_TIMED_OUT
dtx.UpdateTime = _utcnow()
dtx.save(update_fields=['TransactionStatus', 'UpdateTime'])
@staticmethod
def _get_service_domain(step):
"""获取步骤参与服务的域名"""
if step.ParticipantServiceUUID:
return step.ParticipantServiceUUID.ServiceDomain.rstrip('/')
return None
@staticmethod
def _execute_step(step, access_token=None):
"""执行单个步骤——调用参与服务的本地事务管理器 execute 端点"""
step.StepStatus = TransactionStep.STEP_STATUS_RUNNING
step.StartTime = _utcnow()
step.save(update_fields=['StepStatus', 'StartTime'])
service_domain = TransactionManager._get_service_domain(step)
if not service_domain:
raise StepExecutionError(
f'步骤无参与服务或服务域名未配置: step={step.StepIndex}',
step_index=step.StepIndex,
)
endpoint = f'{service_domain}{LOCAL_EXECUTE_PATH}'
payload = {
'StepName': step.StepName,
'payload': step.ActionPayload,
'TransactionCode': step.TransactionUUID.TransactionCode,
'StepIndex': step.StepIndex,
}
return TransactionManager._call_service(
endpoint=endpoint,
payload=payload,
access_token=access_token,
)
@staticmethod
def _compensate_step(step, access_token=None):
"""补偿单个步骤——调用参与服务的本地事务管理器 compensate 端点"""
service_domain = TransactionManager._get_service_domain(step)
if not service_domain:
raise StepExecutionError(
f'步骤无参与服务或服务域名未配置: step={step.StepIndex}',
step_index=step.StepIndex,
)
endpoint = f'{service_domain}{LOCAL_COMPENSATE_PATH}'
payload = {
'StepName': step.StepName,
'payload': step.CompensatePayload,
'ActionResult': step.StepResult,
'TransactionCode': step.TransactionUUID.TransactionCode,
'StepIndex': step.StepIndex,
}
return TransactionManager._call_service(
endpoint=endpoint,
payload=payload,
access_token=access_token,
)
@staticmethod
def _call_service(endpoint, payload=None, access_token=None):
"""调用远程服务的本地事务管理器端点(使用连接池 + TLS 验证)"""
headers = {'Content-Type': 'application/json'}
if access_token:
headers['Authorization'] = f'Service {access_token}'
try:
resp = _http_session.post(
endpoint,
json=payload or {},
headers=headers,
timeout=30,
)
response_data = resp.text
# 非 2xx:无论响应体格式(HTML 错误页/网关页等)一律视为失败,
# 避免步骤被误判为 Completed
if not resp.ok:
raise StepExecutionError(
f'本地事务管理器调用失败: HTTP {resp.status_code}',
status_code=resp.status_code,
response_body=response_data[:4096],
)
content_type = resp.headers.get('Content-Type', '')
if 'application/json' in content_type:
try:
result = resp.json()
except ValueError:
# JSONDecodeError 是 ValueError 子类:响应体声明 JSON 却无法解析时,
# 转 StepExecutionError,保证失败必进补偿链路,事务不悬挂 Running
raise StepExecutionError(
f'本地事务管理器返回非法 JSON: {response_data[:512]}',
status_code=resp.status_code,
response_body=response_data[:4096],
)
if isinstance(result, dict) and result.get('code') == 200:
return result.get('data', result)
elif isinstance(result, dict) and 'success' in result:
if not result['success']:
raise StepExecutionError(
result.get('error', '本地执行失败'),
response_body=response_data,
)
return result.get('result', result)
return result
return {'raw': response_data}
except requests.exceptions.RequestException as e:
# 统一兜底:连接失败/超时/重试耗尽等任何 requests 异常
# 一律转 StepExecutionError,确保失败必进补偿链路
raise StepExecutionError(f'本地事务管理器调用失败: {e}')
+36
View File
@@ -0,0 +1,36 @@
from django.urls import path
from gvsdsdk.msync import views
app_name = 'msync'
urlpatterns = [
path('auth/token/', views.AuthTokenView.as_view(), name='auth_token'),
path('auth/token/refresh/', views.AuthTokenRefreshView.as_view(), name='auth_token_refresh'),
path('auth/introspect/', views.AuthTokenIntrospectView.as_view(), name='auth_introspect'),
path('registry/', views.RegistryListView.as_view(), name='registry_list'),
path('registry/<str:name>/', views.RegistryDetailView.as_view(), name='registry_detail'),
path('registry/<str:name>/heartbeat/', views.RegistryHeartbeatView.as_view(), name='registry_heartbeat'),
path('registry/<str:name>/deregister/', views.RegistryDeregisterView.as_view(), name='registry_deregister'),
path('ogm/', views.OGMRegisterView.as_view(), name='ogm_register'),
path('ogm/batch/', views.OGMBatchRegisterView.as_view(), name='ogm_batch_register'),
path('ogm/query/', views.OGMQueryView.as_view(), name='ogm_query'),
path('ogm/<str:global_id>/', views.OGMResolveView.as_view(), name='ogm_resolve'),
path('ogm/<str:global_id>/update/', views.OGMUpdateView.as_view(), name='ogm_update'),
path('ogm/<str:global_id>/deregister/', views.OGMDeregisterView.as_view(), name='ogm_deregister'),
path('tx/', views.TransactionCreateView.as_view(), name='tx_create'),
path('tx/query/', views.TransactionQueryView.as_view(), name='tx_query'),
path('tx/timeout-check/', views.TransactionTimeoutCheckView.as_view(), name='tx_timeout_check'),
path('tx/<str:transaction_code>/', views.TransactionDetailView.as_view(), name='tx_detail'),
path('tx/<str:transaction_code>/execute/', views.TransactionExecuteView.as_view(), name='tx_execute'),
path('tx/<str:transaction_code>/compensate/', views.TransactionCompensateView.as_view(), name='tx_compensate'),
path('tx/<str:transaction_code>/cancel/', views.TransactionCancelView.as_view(), name='tx_cancel'),
path('tx/<str:transaction_code>/retry/', views.TransactionRetryStepView.as_view(), name='tx_retry_step'),
path('tx/<str:transaction_code>/retry-compensate/', views.TransactionRetryCompensateView.as_view(), name='tx_retry_compensate'),
path('local/execute/', views.LocalExecuteView.as_view(), name='local_execute'),
path('local/compensate/', views.LocalCompensateView.as_view(), name='local_compensate'),
path('local/handlers/', views.LocalHandlersView.as_view(), name='local_handlers'),
]
+537
View File
@@ -0,0 +1,537 @@
import logging
import traceback
from rest_framework.views import APIView
from rest_framework.response import Response
from rest_framework.permissions import IsAuthenticated
from gvsdsdk.msync.auth import (
AuthenticateService,
RefreshServiceToken,
ValidateAccessToken,
)
from gvsdsdk.msync.middleware import ServiceTokenAuthentication
from gvsdsdk.msync.registry import ServiceRegistry, RegistryError
from gvsdsdk.msync.ogm import OGMManager, OGMError
from gvsdsdk.msync.transaction import TransactionManager, TransactionError
from gvsdsdk.msync.local_tx import local_tx
from gvsdsdk.msync.serializers import (
AuthTokenRequestSerializer,
RefreshTokenRequestSerializer,
ServiceRegisterSerializer,
ServiceNodeSerializer,
OGMRegisterSerializer,
OGMBatchRegisterSerializer,
ServiceTokenResponseSerializer,
TransactionCreateSerializer,
TransactionExecuteSerializer,
TransactionRetryStepSerializer,
LocalExecuteSerializer,
LocalCompensateSerializer,
)
logger = logging.getLogger(__name__)
def _ok(data=None, msg='success'):
return {'code': 200, 'msg': msg, 'data': data}
def _err(msg, code=400):
return {'code': code, 'msg': msg, 'data': None}
class AuthTokenView(APIView):
authentication_classes = []
permission_classes = []
def post(self, request):
try:
ser = AuthTokenRequestSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
token = AuthenticateService(
service_name=ser.ValidatedData['ServiceName'],
signed_payload=ser.ValidatedData['Payload'],
signature=ser.ValidatedData['Signature'],
)
if token is None:
return Response(_err('认证失败', 401), status=401)
try:
scope = token.TokenScope
if not isinstance(scope, list):
scope = list(scope) if scope else []
except Exception:
scope = []
try:
expires_ts = token.ExpiresAt.timestamp()
except Exception:
import time as _time
expires_ts = _time.time() + 7200
data = ServiceTokenResponseSerializer({
'AccessToken': token.AccessToken,
'RefreshToken': token.RefreshToken,
'Scope': scope,
'ExpiresAt': expires_ts,
}).data
return Response(_ok(data))
except Exception as e:
logger.error(f'认证令牌签发异常: {e}\n{traceback.format_exc()}')
return Response(_err('服务器内部错误', 500), status=500)
class AuthTokenRefreshView(APIView):
authentication_classes = []
permission_classes = []
def post(self, request):
try:
ser = RefreshTokenRequestSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
token = RefreshServiceToken(ser.ValidatedData['RefreshToken'])
if token is None:
return Response(_err('刷新失败', 401), status=401)
try:
scope = token.TokenScope
if not isinstance(scope, list):
scope = list(scope) if scope else []
except Exception:
scope = []
try:
expires_ts = token.ExpiresAt.timestamp()
except Exception:
import time as _time
expires_ts = _time.time() + 7200
data = ServiceTokenResponseSerializer({
'AccessToken': token.AccessToken,
'RefreshToken': token.RefreshToken,
'Scope': scope,
'ExpiresAt': expires_ts,
}).data
return Response(_ok(data))
except Exception as e:
logger.error(f'令牌刷新异常: {e}\n{traceback.format_exc()}')
return Response(_err('服务器内部错误', 500), status=500)
class AuthTokenIntrospectView(APIView):
"""Token 自省端点 — 供子服务器验证 ServiceToken 有效性
子服务器无法直接访问主服务器的 Django ORM 和 MASTER_SECRET,
因此通过此端点远程验证 token。
"""
authentication_classes = []
permission_classes = []
def post(self, request):
access_token = request.data.get('AccessToken', '')
if not access_token:
return Response(_err('AccessToken 不能为空', 400), status=400)
try:
payload = ValidateAccessToken(access_token)
if payload is None:
return Response({'Valid': False, 'Active': False})
return Response({'Valid': True, 'Active': True, 'Payload': payload})
except Exception as e:
logger.error(f'Token 自省异常: {e}\n{traceback.format_exc()}')
return Response({'Valid': False, 'Active': False})
class RegistryListView(APIView):
authentication_classes = [ServiceTokenAuthentication]
def get(self, request):
if not request.user.is_authenticated:
return Response(_err('需要服务认证', 401), status=401)
services = ServiceRegistry.DiscoverAll(active_only=True)
return Response(_ok(services))
def post(self, request):
# 注册端点:新服务尚未拥有令牌,无需认证,但必须携带密钥
# TODO: 考虑添加速率限制以防止滥用注册
ser = ServiceRegisterSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
node = ServiceRegistry.Register(
name=ser.ValidatedData['ServiceName'],
domain=ser.ValidatedData['ServiceDomain'],
secret_hash=ser.ValidatedData.get('ServiceSecretHash', ''),
raw_secret=ser.ValidatedData.get('ServiceSecret'),
scope=ser.ValidatedData.get('ServiceScope'),
service_type=ser.ValidatedData.get('ServiceType', 0),
metadata=ser.ValidatedData.get('ServiceMetadata', {}),
)
except RegistryError as e:
return Response(_err(str(e), 409), status=409)
data = ServiceNodeSerializer(node).data
return Response(_ok(data), status=201)
class RegistryDetailView(APIView):
authentication_classes = [ServiceTokenAuthentication]
def get(self, request, name):
if not request.user.is_authenticated:
return Response(_err('需要服务认证', 401), status=401)
info = ServiceRegistry.Discover(name)
if info is None:
return Response(_err('服务不存在', 404), status=404)
return Response(_ok(info))
class RegistryHeartbeatView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request, name):
# 仅允许持有该服务密钥签发的令牌执行本服务心跳
if getattr(request, 'service_name', None) != name:
return Response(_err('无权操作其他服务', 403), status=403)
try:
ServiceRegistry.Heartbeat(name)
except RegistryError as e:
return Response(_err(str(e), 404), status=404)
return Response(_ok(msg='心跳成功'))
class RegistryDeregisterView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def delete(self, request, name):
# 仅允许持有该服务密钥签发的令牌注销本服务
if getattr(request, 'service_name', None) != name:
return Response(_err('无权操作其他服务', 403), status=403)
try:
ServiceRegistry.Deregister(name)
except RegistryError as e:
return Response(_err(str(e), 404), status=404)
return Response(_ok(msg='注销成功'))
class OGMRegisterView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request):
ser = OGMRegisterSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
ogm = OGMManager.Register(
object_type=ser.ValidatedData['ObjectType'],
local_id=ser.ValidatedData['ObjectLocalID'],
service_name=ser.ValidatedData['ServiceName'],
global_id=ser.ValidatedData.get('ObjectGlobalID'),
metadata=ser.ValidatedData.get('Metadata', {}),
)
except OGMError as e:
return Response(_err(str(e), 409), status=409)
return Response(_ok(OGMManager._to_dict(ogm)), status=201)
class OGMBatchRegisterView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request):
ser = OGMBatchRegisterSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
results = OGMManager.RegisterBatch(
items=ser.ValidatedData['Items'],
service_name=ser.ValidatedData['ServiceName'],
)
except OGMError as e:
return Response(_err(str(e), 409), status=409)
data = [OGMManager._to_dict(ogm) for ogm in results]
return Response(_ok(data), status=201)
class OGMResolveView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def get(self, request, global_id):
info = OGMManager.Resolve(global_id)
if info is None:
return Response(_err('映射不存在', 404), status=404)
return Response(_ok(info))
class OGMQueryView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def get(self, request):
ObjectType = request.query_params.get('ObjectType')
ServiceName = request.query_params.get('ServiceName')
LocalID = request.query_params.get('LocalID')
if LocalID and ObjectType and ServiceName:
info = OGMManager.ResolveLocal(ObjectType, LocalID, ServiceName)
if info is None:
return Response(_err('映射不存在', 404), status=404)
return Response(_ok(info))
if ObjectType:
results = OGMManager.QueryByType(ObjectType)
return Response(_ok(results))
if ServiceName:
results = OGMManager.QueryByService(ServiceName, object_type=ObjectType)
return Response(_ok(results))
return Response(_err('请提供 ObjectType 或 ServiceName 参数', 400), status=400)
class OGMUpdateView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def put(self, request, global_id):
metadata = request.data.get('Metadata')
ogm = OGMManager.Update(global_id, metadata=metadata)
if ogm is None:
return Response(_err('映射不存在', 404), status=404)
return Response(_ok(OGMManager._to_dict(ogm)))
class OGMDeregisterView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def delete(self, request, global_id):
success = OGMManager.Deregister(global_id)
if not success:
return Response(_err('映射不存在', 404), status=404)
return Response(_ok(msg='注销成功'))
class TransactionCreateView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request):
ser = TransactionCreateSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
dtx = TransactionManager.Create(
name=ser.ValidatedData['TransactionName'],
steps_data=ser.ValidatedData['Steps'],
description=ser.ValidatedData.get('TransactionDescription', ''),
transaction_type=ser.ValidatedData.get('TransactionType', 'saga'),
initiator_service_name=ser.ValidatedData.get('InitiatorServiceName'),
reference_ogm_uuid=ser.ValidatedData.get('ReferenceOGMUUID'),
payload=ser.ValidatedData.get('Payload'),
timeout_seconds=ser.ValidatedData.get('TimeoutSeconds', 300),
)
except TransactionError as e:
return Response(_err(str(e), 400), status=400)
return Response(_ok(TransactionManager.GetDetail(dtx.TransactionCode)), status=201)
class TransactionExecuteView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request, transaction_code):
ser = TransactionExecuteSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
dtx = TransactionManager.Execute(
transaction_code,
access_token=ser.ValidatedData.get('AccessToken'),
)
except TransactionError as e:
return Response(_err(str(e), 400), status=400)
return Response(_ok(TransactionManager.GetDetail(dtx.TransactionCode)))
class TransactionCompensateView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request, transaction_code):
access_token = request.data.get('AccessToken')
try:
dtx = TransactionManager.Compensate(
transaction_code,
access_token=access_token,
)
except TransactionError as e:
return Response(_err(str(e), 400), status=400)
return Response(_ok(TransactionManager.GetDetail(dtx.TransactionCode)))
class TransactionCancelView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request, transaction_code):
try:
TransactionManager.Cancel(transaction_code)
except TransactionError as e:
return Response(_err(str(e), 400), status=400)
return Response(_ok(msg='事务已取消'))
class TransactionDetailView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def get(self, request, transaction_code):
detail = TransactionManager.GetDetail(transaction_code)
if detail is None:
return Response(_err('事务不存在', 404), status=404)
return Response(_ok(detail))
class TransactionRetryStepView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request, transaction_code):
ser = TransactionRetryStepSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
step = TransactionManager.RetryStep(
transaction_code,
step_index=ser.ValidatedData['StepIndex'],
access_token=ser.ValidatedData.get('AccessToken'),
)
except TransactionError as e:
return Response(_err(str(e), 400), status=400)
return Response(_ok({
'StepIndex': step.StepIndex,
'StepName': step.StepName,
'Status': step.get_StepStatus_display(),
'RetryCount': step.RetryCount,
}))
class TransactionRetryCompensateView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request, transaction_code):
ser = TransactionRetryStepSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
try:
step = TransactionManager.RetryCompensate(
transaction_code,
step_index=ser.ValidatedData['StepIndex'],
access_token=ser.ValidatedData.get('AccessToken'),
)
except TransactionError as e:
return Response(_err(str(e), 400), status=400)
return Response(_ok({
'StepIndex': step.StepIndex,
'StepName': step.StepName,
'Status': step.get_StepStatus_display(),
'RetryCount': step.RetryCount,
'TransactionStatus': step.TransactionUUID.get_TransactionStatus_display(),
}))
class TransactionTimeoutCheckView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request):
timed_out = TransactionManager.CheckTimeouts()
return Response(_ok({'TimedOut': timed_out, 'Count': len(timed_out)}))
class TransactionQueryView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def get(self, request):
status = request.query_params.get('status')
if status is None:
return Response(_err('请提供 status 参数', 400), status=400)
try:
status_int = int(status)
except ValueError:
return Response(_err('status 必须为整数', 400), status=400)
results = TransactionManager.QueryByStatus(status_int)
return Response(_ok(results))
class LocalExecuteView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request):
ser = LocalExecuteSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
result = local_tx.ExecuteStep(
name=ser.ValidatedData['StepName'],
payload=ser.ValidatedData.get('Payload'),
)
if result['success']:
return Response(_ok(result['result']))
return Response(_err(result['error'], 500), status=500)
class LocalCompensateView(APIView):
authentication_classes = [ServiceTokenAuthentication]
permission_classes = [IsAuthenticated]
def post(self, request):
ser = LocalCompensateSerializer(data=request.data)
if not ser.IsValid():
return Response(_err(ser.errors, 400), status=400)
result = local_tx.CompensateStep(
name=ser.ValidatedData['StepName'],
payload=ser.ValidatedData.get('Payload'),
action_result=ser.ValidatedData.get('ActionResult'),
)
if result['success']:
return Response(_ok(result['result']))
return Response(_err(result['error'], 500), status=500)
class LocalHandlersView(APIView):
def get(self, request):
handlers = local_tx.ListHandlers()
return Response(_ok(handlers))
+66
View File
@@ -0,0 +1,66 @@
"""GVSDS SDK 配置与密钥访问
环境变量优先,其次读取项目根目录 secrets.ini。
与 gvsds.secrets 同源(同一 secrets.ini),供 SDK 在无 gvsds 依赖时使用。
env 键规则:`{SECTION}_{KEY}` 全大写(如 MSYNC_MASTER_SECRET)。
"""
import os
import configparser
from pathlib import Path
_SECRETS_FILE = Path(__file__).resolve().parent.parent / "secrets.ini"
_config = None
def _load_config():
global _config
if _config is not None:
return _config
_config = configparser.ConfigParser()
if _SECRETS_FILE.exists():
_config.read(str(_SECRETS_FILE), encoding="utf-8")
return _config
def get(section, key, fallback=None):
cfg = _load_config()
env_key = f"{section.upper()}_{key.upper()}"
env_val = os.environ.get(env_key)
if env_val is not None:
return env_val
return cfg.get(section, key, fallback=fallback)
def getint(section, key, fallback=0):
val = get(section, key, fallback=None)
if val is None:
return fallback
try:
return int(val)
except (ValueError, TypeError):
return fallback
def getfloat(section, key, fallback=0.0):
val = get(section, key, fallback=None)
if val is None:
return fallback
try:
return float(val)
except (ValueError, TypeError):
return fallback
def getboolean(section, key, fallback=False):
val = get(section, key, fallback=None)
if val is None:
return fallback
return val.lower() in ("true", "1", "yes", "on")
def reload():
global _config
_config = None
_load_config()
+26
View File
@@ -0,0 +1,26 @@
"""GVSDS SDK DRF 序列化器基类
提供 PascalCase 命名风格的 DRF 序列化器基类(is_valid/validated_data 的
PascalCase 别名)。gvsds.common.Serializer 从本模块 re-export。"""
from rest_framework import serializers
class PascalCaseSerializer(serializers.Serializer):
def IsValid(self, raise_exception=False):
return self.is_valid(raise_exception=raise_exception)
@property
def ValidatedData(self):
return self.validated_data
class PascalCaseModelSerializer(serializers.ModelSerializer):
def IsValid(self, raise_exception=False):
return self.is_valid(raise_exception=raise_exception)
@property
def ValidatedData(self):
return self.validated_data