main.go 6.33 KB
Newer Older
1
2
3
package web

import (
liming6's avatar
liming6 committed
4
	"encoding/base64"
5
	"get-container/cmd/opsflow/backend"
liming6's avatar
liming6 committed
6
	"net/http"
7
	"strings"
liming6's avatar
liming6 committed
8
	"time"
9
10

	"github.com/gin-gonic/gin"
liming6's avatar
liming6 committed
11
12
	"github.com/pquerna/otp"
	"github.com/pquerna/otp/totp"
13
	"github.com/spf13/viper"
14
15
16

	swaggerFiles "github.com/swaggo/files"
	ginSwagger "github.com/swaggo/gin-swagger"
17
18
)

19
20
// swagger embed files

21
22
23
24
25
26
27
28
var (
	globalCfg *viper.Viper = nil
)

func Init(cfg *viper.Viper) {
	globalCfg = cfg
}

29
30
31
32
33
34
35
type RestfulResult[T any] struct {
	Code int    `json:"code"`
	Msg  string `json:"msg"`
	Data T      `json:"data,omitempty"`
}

type RestfulNoDataResult struct {
36
37
	Code int    `json:"code"`
	Msg  string `json:"msg"`
38
39
40
41
42
43
}

type RestfulListResult[T any] struct {
	Code int    `json:"code"`
	Msg  string `json:"msg"`
	Data []T    `json:"data,omitempty"`
44
45
46
47
}

func ReturnGin(ctx *gin.Context, data any, err error) {
	if err != nil {
48
		ctx.JSON(500, RestfulNoDataResult{
49
50
51
52
53
			Code: 500,
			Msg:  err.Error(),
		})
		return
	}
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
	ctx.JSON(200, RestfulResult[any]{
		Code: 200,
		Msg:  "ok",
		Data: data,
	})
}

func ReturnGinList[T any](ctx *gin.Context, data []T, err error) {
	if err != nil {
		ctx.JSON(500, RestfulNoDataResult{
			Code: 500,
			Msg:  err.Error(),
		})
		return
	}
	ctx.JSON(200, RestfulListResult[T]{
70
71
72
73
74
75
		Code: 200,
		Msg:  "ok",
		Data: data,
	})
}

liming6's avatar
liming6 committed
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
func webAuth(ctx *gin.Context) {
	authHeader := ctx.GetHeader("Authorization")
	if authHeader == "" {
		ctx.JSON(http.StatusUnauthorized, gin.H{"error": "Not find header Authorization"})
		ctx.Abort()
		return
	}
	fields := strings.SplitN(authHeader, " ", 2)
	if len(fields) != 2 || fields[0] != "Bearer" {
		ctx.JSON(http.StatusUnauthorized, gin.H{"error": "Authorization error"})
		ctx.Abort()
		return
	}
	code, err := base64.StdEncoding.DecodeString(fields[1])
	if err != nil {
		ctx.JSON(http.StatusUnauthorized, gin.H{"error": err.Error()})
		ctx.Abort()
		return
	}
	sec := globalCfg.GetString("auth_key")
	ok, err := totp.ValidateCustom(string(code), sec, time.Now(), totp.ValidateOpts{
		Period:    30, // 每 30 秒更新一次
		Skew:      1,  // 允许前后偏移 1 个周期(即允许 30 秒的时间误差)
		Digits:    otp.DigitsSix,
		Algorithm: otp.AlgorithmSHA1,
	})
	if err != nil {
		ctx.JSON(http.StatusUnauthorized, gin.H{"error": err.Error()})
		ctx.Abort()
		return
	}
	if ok {
		ctx.Next()
	} else {
		ctx.JSON(http.StatusUnauthorized, gin.H{"error": "Authorization failed"})
		ctx.Abort()
		return
	}
}

116
117
func WebServer(addr string) error {
	engine := gin.Default()
liming6's avatar
liming6 committed
118
	cmdGroup := engine.Group("/api/cmd")
119
	if globalCfg.GetBool("debug_mode") {
liming6's avatar
liming6 committed
120
		// 调试模式
121
		engine.GET("/swagger/*any", ginSwagger.WrapHandler(swaggerFiles.Handler))
liming6's avatar
liming6 committed
122
123
124
	} else {
		// 非调试模式,添加认证
		cmdGroup.Use(webAuth)
125
	}
liming6's avatar
liming6 committed
126

127
128
129
130
131
132
	cmdGroup.GET("/all", _controller.GetAllInfo)
	cmdGroup.GET("/loginUser", _controller.GetOnlineUser)
	cmdGroup.GET("/sysload", _controller.GetSysLoad)
	cmdGroup.GET("/dcuload", _controller.GetDCULoad)
	cmdGroup.GET("/rcclinfo", _controller.GetRcclHandler)
	cmdGroup.POST("/rccl/post", _controller.PostRcclHandler)
133
134
	return engine.Run(addr)
}
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248

type controller struct{}

var _controller = controller{}

// GetRcclHandler godoc
// @Summary 获取 rccl all_reduce_perf 性能信息
// @Description 获取 rccl all_reduce_perf 性能信息
// @Accept json
// @Produce json
// @Success 200 {object} RestfulResult[backend.RcclTestAllReducePrefResult]
// @Failure 500 {object} RestfulNoDataResult
// @Router /rcclinfo [get]
func (c controller) GetRcclHandler(ctx *gin.Context) {
	r, err := backend.AllReducePerf(globalCfg.GetString("rccl_test_path"), globalCfg.GetString("rccl_all_reduce_perf_args"))
	ReturnGin(ctx, r, err)
}

// GetDCULoad godoc
// @Summary 获取 DCU 负载信息
// @Description 获取 DCU 负载信息
// @Accept json
// @Produce json
// @Success 200 {object} RestfulListResult[backend.DCULoad]
// @Failure 500 {object} RestfulNoDataResult
// @Router /dcuload [get]
func (c controller) GetDCULoad(ctx *gin.Context) {
	dcu, err := backend.GetDCULoad()
	ReturnGinList(ctx, dcu, err)
}

// GetOnlineUser godoc
// @Summary 获取在线用户信息
// @Description 获取在线用户信息
// @Accept json
// @Produce json
// @Success 200 {object} RestfulListResult[backend.LoginUserInfo]
// @Failure 500 {object} RestfulNoDataResult
// @Router /loginUser [get]
func (c controller) GetOnlineUser(ctx *gin.Context) {
	olu, err := backend.GetOnlineUser()
	ReturnGinList(ctx, olu, err)
}

// GetSysLoad godoc
// @Summary 获取系统负载信息
// @Description 获取系统负载信息
// @Accept json
// @Produce json
// @Success 200 {object} RestfulResult[backend.SysInfo]
// @Failure 500 {object} RestfulNoDataResult
// @Router /sysload [get]
func (c controller) GetSysLoad(ctx *gin.Context) {
	sys, err := backend.GetSysLoad()
	ReturnGin(ctx, sys, err)
}

// GetAllInfo godoc
// @Summary 获取所有信息(系统负载、DCU 负载、在线用户)
// @Description 获取所有信息(系统负载、DCU 负载、在线用户)
// @Accept json
// @Produce json
// @Success 200 {object} RestfulResult[backend.AllInfo]
// @Failure 500 {object} RestfulNoDataResult
// @Router /all [get]
func (c controller) GetAllInfo(ctx *gin.Context) {
	olu, err := backend.GetOnlineUser()
	if err != nil {
		ReturnGin(ctx, nil, err)
	}
	sys, err := backend.GetSysLoad()
	if err != nil {
		ReturnGin(ctx, nil, err)
	}
	dcu, err := backend.GetDCULoad()
	if err != nil {
		ReturnGin(ctx, nil, err)
	}
	ReturnGin(ctx, backend.AllInfo{
		DCUInfo:        dcu,
		SysInfo:        *sys,
		OnlineUserInfo: olu,
	}, err)
}

type RcclArgs struct {
	Args []string `json:"args"`
}

// PostRcclHandler godoc
// @Summary 给出rccl all_reduce_perf参数,执行单机测试
// @Description 给出rccl all_reduce_perf参数,执行单机测试
// @Accept json
// @Produce json
// @Param   args body  RcclArgs true "rccl all reduce perf args"
// @Success 200 {object} RestfulResult[backend.RcclTestAllReducePrefResult]
// @Failure 500 {object} RestfulNoDataResult
// @Router /rccl/post [post]
func (c controller) PostRcclHandler(ctx *gin.Context) {
	args := RcclArgs{}
	err := ctx.BindJSON(&args)
	if err != nil {
		ReturnGin(ctx, nil, err)
		return
	}
	if len(args.Args) == 0 {
		r, err := backend.AllReducePerf(globalCfg.GetString("rccl_test_path"), globalCfg.GetString("rccl_all_reduce_perf_args"))
		ReturnGin(ctx, r, err)
		return
	}
	arg := strings.Join(args.Args, " ")
	r, err := backend.AllReducePerf(globalCfg.GetString("rccl_test_path"), arg)
	ReturnGin(ctx, r, err)
}