main.go 5.04 KB
Newer Older
1
2
3
4
package web

import (
	"get-container/cmd/opsflow/backend"
5
	"strings"
6
7

	"github.com/gin-gonic/gin"
8
	"github.com/spf13/viper"
9
10
11

	swaggerFiles "github.com/swaggo/files"
	ginSwagger "github.com/swaggo/gin-swagger"
12
13
)

14
15
// swagger embed files

16
17
18
19
20
21
22
23
var (
	globalCfg *viper.Viper = nil
)

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

24
25
26
27
28
29
30
type RestfulResult[T any] struct {
	Code int    `json:"code"`
	Msg  string `json:"msg"`
	Data T      `json:"data,omitempty"`
}

type RestfulNoDataResult struct {
31
32
	Code int    `json:"code"`
	Msg  string `json:"msg"`
33
34
35
36
37
38
}

type RestfulListResult[T any] struct {
	Code int    `json:"code"`
	Msg  string `json:"msg"`
	Data []T    `json:"data,omitempty"`
39
40
41
42
}

func ReturnGin(ctx *gin.Context, data any, err error) {
	if err != nil {
43
		ctx.JSON(500, RestfulNoDataResult{
44
45
46
47
48
			Code: 500,
			Msg:  err.Error(),
		})
		return
	}
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
	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]{
65
66
67
68
69
70
71
72
		Code: 200,
		Msg:  "ok",
		Data: data,
	})
}

func WebServer(addr string) error {
	engine := gin.Default()
73
74
75
	if globalCfg.GetBool("debug_mode") {
		engine.GET("/swagger/*any", ginSwagger.WrapHandler(swaggerFiles.Handler))
	}
76
	cmdGroup := engine.Group("/api/cmd")
77
78
79
80
81
82
	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)
83
84
	return engine.Run(addr)
}
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
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
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

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)
}