1、初始化

This commit is contained in:
wangjian 2023-03-23 15:58:12 +08:00
commit 46d29e280b
24 changed files with 1371 additions and 0 deletions

1
README.md Normal file
View File

@ -0,0 +1 @@
## 任务调度中心

154
cmd/server.go Normal file
View File

@ -0,0 +1,154 @@
package cmd
import (
"context"
"fmt"
"hpds_control_center/model"
"hpds_control_center/mq"
"os"
"os/signal"
"syscall"
"git.hpds.cc/Component/logging"
"github.com/spf13/cobra"
"go.uber.org/zap"
"hpds_control_center/config"
discover "hpds_control_center/pkg/discover/consul"
)
var (
ConfigFileFlag string = "./config/config.yaml"
ConsulAddress string = "http://localhost:8500"
NodeName string = "main-node"
Mode string = "dev"
)
func must(err error) {
if err != nil {
_, _ = fmt.Fprint(os.Stderr, err)
os.Exit(1)
}
}
func NewStartCmd() *cobra.Command {
cmd := &cobra.Command{
Use: "start",
Short: "Start hpds_web application",
Run: func(cmd *cobra.Command, args []string) {
var (
cfg *config.WebConfig
err error
)
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
must(err)
configFileFlag, err := cmd.Flags().GetString("c")
if err != nil {
fmt.Println("get local config err: ", err)
return
}
ConsulAddress, err = cmd.Flags().GetString("r")
if err != nil {
fmt.Println("get remote config err: ", err)
return
}
NodeName, err = cmd.Flags().GetString("n")
if err != nil {
fmt.Println("get remote path config err: ", err)
return
}
Mode, err = cmd.Flags().GetString("m")
if err != nil {
fmt.Println("get remote path config err: ", err)
return
}
if len(configFileFlag) > 1 {
cfg, err = config.ParseConfigByFile(configFileFlag)
must(err)
err = config.UpdateRemoteConfig(cfg)
must(err)
ConfigFileFlag = configFileFlag
} else {
//获取consul注册中心的配置文件
cfg, err = config.GetRemoteConfig(ConsulAddress, fmt.Sprintf("hpds-pavement/hpds_control/%s/%s", Mode, NodeName))
must(err)
err = config.UpdateLocalConfig(cfg, ConfigFileFlag)
}
//创建注册对象
tags := make([]string, 1)
tags[0] = "control"
consulCfg, err := discover.NewConsulConfig(cfg.Consul.Host, cfg.Name, cfg.Name, cfg.Consul.Host, cfg.Consul.Port,
tags, 300, 300, 300)
must(err)
//连接数据库
model.New(cfg.Db.DriveName, cfg.Db.Conn, cfg.Mode == "dev")
logger := LoadLoggerConfig(cfg.Logging)
//创建消息连接点
mq.MqList, err = mq.NewMqClient(cfg.Funcs, cfg.Node, logger)
must(err)
// 退出channel
exitChannel := make(chan os.Signal)
defer close(exitChannel)
// 退出信号监听
go func(c chan os.Signal) {
_ = consulCfg.ServiceDeregister()
signal.Notify(c, syscall.SIGINT, syscall.SIGTERM)
}(exitChannel)
// start http service
go func() {
//fmt.Printf("Http Server start at port %d \n", cfg.Port)
//启动前执行注册
err = consulCfg.ServiceRegister()
must(err)
}()
//服务退出取消注册
//err = consulCfg.ServiceDeregister()
//
//must(err)
//zap.L().Error("发生错误", zap.Error(err))
for {
select {
case <-ctx.Done():
_ = consulCfg.ServiceDeregister()
logger.With(
zap.String("web", "exit"),
).Error(ctx.Err().Error())
return
case errs := <-exitChannel:
_ = consulCfg.ServiceDeregister()
logger.With(
zap.String("web", "服务退出"),
).Info(errs.String())
return
}
}
},
}
cmd.Flags().StringVar(&ConfigFileFlag, "c", "./config/config.yaml", "The configuration file path")
cmd.Flags().StringVar(&ConsulAddress, "r", "http://consul.hpds.cc", "The configuration remote consul address")
cmd.Flags().StringVar(&NodeName, "n", "main-node", "The configuration name")
cmd.Flags().StringVar(&Mode, "m", "dev", "run mode : dev | test | releases")
return cmd
}
func LoadLoggerConfig(opt config.LogOptions) *logging.Logger {
return logging.NewLogger(
logging.SetPath(opt.Path),
logging.SetPrefix(opt.Prefix),
logging.SetDevelopment(opt.Development),
logging.SetDebugFileSuffix(opt.DebugFileSuffix),
logging.SetWarnFileSuffix(opt.WarnFileSuffix),
logging.SetErrorFileSuffix(opt.ErrorFileSuffix),
logging.SetInfoFileSuffix(opt.InfoFileSuffix),
logging.SetMaxAge(opt.MaxAge),
logging.SetMaxBackups(opt.MaxBackups),
logging.SetMaxSize(opt.MaxSize),
logging.SetLevel(logging.LogLevel["debug"]),
)
}

43
config/config-dev.yaml Normal file
View File

@ -0,0 +1,43 @@
name: control_center
host: 0.0.0.0
port: 8088
mode: dev
logging:
path: ./logs
prefix: hpds-control
errorFileSuffix: error.log
warnFileSuffix: warn.log
infoFileSuffix: info.log
debugFileSuffix: debug.log
maxSize: 100
maxBackups: 3000
maxAge: 30
development: true
consul:
host: http://consul.hpds.cc
port: 80
interval: 300
timeout: 5
deregister: 1
db:
conn: root:123456@tcp(127.0.0.1:3306)/hpds_jky?charset=utf8mb4
drive_name: mysql
cache:
host: 127.0.0.1
port: 6379
db: 0
pool_size: 10
node:
host: 127.0.0.1
port: 27188
token: 06d36c6f5705507dae778fdce90d0767
functions:
- name: task-request
dataTag: 12
mqType: 2
- name: task-response
dataTag: 14
mqType: 1
- name: task-execute
dataTag: 16
mqType: 1

137
config/config.go Normal file
View File

@ -0,0 +1,137 @@
package config
import (
"fmt"
"os"
consulapi "github.com/hashicorp/consul/api"
"github.com/spf13/viper"
_ "github.com/spf13/viper/remote"
yaml "gopkg.in/yaml.v3"
)
type WebConfig struct {
Name string `yaml:"name,omitempty"`
Host string `yaml:"host,omitempty"`
Port int `yaml:"port,omitempty"`
Mode string `yaml:"mode,omitempty"`
Consul ConsulConfig `yaml:"consul,omitempty"`
Db DbConfig `yaml:"db"`
Cache CacheConfig `yaml:"cache"`
Logging LogOptions `yaml:"logging"`
Minio MinioConfig `yaml:"minio"`
Node HpdsNode `yaml:"node,omitempty"`
Funcs []FuncConfig `yaml:"functions,omitempty"`
}
type ConsulConfig struct {
Host string `yaml:"host,omitempty"`
Port int `yaml:"port,omitempty"`
Interval int `yaml:"interval,omitempty"`
Timeout int `yaml:"timeout,omitempty"`
Deregister int `yaml:"deregister,omitempty"`
Tags []string `yaml:"tags,omitempty"`
}
type DbConfig struct {
Conn string `yaml:"conn"`
DriveName string `yaml:"drive_name"`
}
type CacheConfig struct {
Host string `yaml:"host,omitempty"`
Port int `yaml:"port,omitempty"`
Pass string `yaml:"pass,omitempty"`
DB int `yaml:"db,omitempty"`
PoolSize int `yaml:"pool_size,omitempty"`
}
type LogOptions struct {
Path string `yaml:"path" json:"path" toml:"path"` // 文件保存地方
Prefix string `yaml:"prefix" json:"prefix" toml:"prefix"` // 日志文件前缀
ErrorFileSuffix string `yaml:"errorFileSuffix" json:"errorFileSuffix" toml:"errorFileSuffix"` // error日志文件后缀
WarnFileSuffix string `yaml:"warnFileSuffix" json:"warnFileSuffix" toml:"warnFileSuffix"` // warn日志文件后缀
InfoFileSuffix string `yaml:"infoFileSuffix" json:"infoFileSuffix" toml:"infoFileSuffix"` // info日志文件后缀
DebugFileSuffix string `yaml:"debugFileSuffix" json:"debugFileSuffix" toml:"debugFileSuffix"` // debug日志文件后缀
Level string `yaml:"level" json:"level" toml:"level"` // 日志等级
MaxSize int `yaml:"maxSize" json:"maxSize" toml:"maxSize"` // 日志文件大小M
MaxBackups int `yaml:"maxBackups" json:"maxBackups" toml:"maxBackups"` // 最多存在多少个切片文件
MaxAge int `yaml:"maxAge" json:"maxAge" toml:"maxAge"` // 保存的最大天数
Development bool `yaml:"development" json:"development" toml:"development"` // 是否是开发模式
}
type MinioConfig struct {
Protocol string `yaml:"protocol"` //http or https
Endpoint string `yaml:"endpoint"`
AccessKeyId string `yaml:"accessKeyId"`
SecretAccessKey string `yaml:"secretAccessKey"`
Bucket string `yaml:"bucket"`
}
type HpdsNode struct {
Host string `yaml:"host"`
Port int `yaml:"port"`
Token string `yaml:"token,omitempty"`
}
type FuncConfig struct {
Name string `yaml:"name"`
DataTag uint8 `yaml:"dataTag"`
MqType uint `yaml:"mqType"` //消息类型, 发布1订阅2
}
func ParseConfigByFile(path string) (cfg *WebConfig, err error) {
buffer, err := os.ReadFile(path)
if err != nil {
return nil, err
}
return load(buffer)
}
func load(buf []byte) (cfg *WebConfig, err error) {
cViper := viper.New()
cViper.SetConfigType("yaml")
cfg = new(WebConfig)
cfg.Funcs = make([]FuncConfig, 0)
//cViper.ReadConfig(bytes.NewBuffer(buf))
err = yaml.Unmarshal(buf, cfg)
//err = cViper.Unmarshal(cfg)
//if err != nil {
// return nil, err
//}
return
}
func UpdateLocalConfig(cfg *WebConfig, fn string) error {
data, err := yaml.Marshal(cfg)
if err != nil {
return err
}
err = os.WriteFile(fn, data, 0600)
return err
}
func UpdateRemoteConfig(cfg *WebConfig) error {
consulClient, err := consulapi.NewClient(&consulapi.Config{Address: fmt.Sprintf("%s:%d", cfg.Consul.Host, cfg.Consul.Port)})
if err != nil {
return err
}
val, err := yaml.Marshal(cfg)
if err != nil {
return err
}
p := &consulapi.KVPair{Key: fmt.Sprintf("hpds-pavement/hpds_control_center/%s/%s", cfg.Mode, cfg.Name), Value: val}
if _, err = consulClient.KV().Put(p, nil); err != nil {
return err
}
return nil
}
func GetRemoteConfig(remoteAddr, path string) (cfg *WebConfig, err error) {
consulClient, err := consulapi.NewClient(&consulapi.Config{Address: remoteAddr})
if err != nil {
return nil, err
}
kv, _, err := consulClient.KV().Get(path, nil)
if err != nil {
return nil, err
}
return load(kv.Value)
}

48
config/config.yaml Normal file
View File

@ -0,0 +1,48 @@
name: control_center
host: 0.0.0.0
port: 8088
mode: dev
logging:
path: ./logs
prefix: hpds-control
errorFileSuffix: error.log
warnFileSuffix: warn.log
infoFileSuffix: info.log
debugFileSuffix: debug.log
maxSize: 100
maxBackups: 3000
maxAge: 30
development: true
consul:
host: http://consul.hpds.cc
port: 80
interval: 300
timeout: 5
deregister: 1
db:
conn: root:123456@tcp(127.0.0.1:3306)/hpds_jky?charset=utf8mb4
drive_name: mysql
cache:
host: 127.0.0.1
port: 6379
db: 0
pool_size: 10
minio:
protocol: http
endpoint: 127.0.0.1:9000
accessKeyId: root
secretAccessKey: OIxv7QptYBO3
node:
host: 114.55.236.153
port: 9188
token: 06d36c6f5705507dae778fdce90d0767
functions:
- name: task-request
dataTag: 12
mqType: 2
- name: task-response
dataTag: 14
mqType: 2
- name: task-execute
dataTag: 16
mqType: 1

102
go.mod Normal file
View File

@ -0,0 +1,102 @@
module hpds_control_center
go 1.19
require (
git.hpds.cc/Component/logging v0.0.0-20230106105738-e378e873921b
git.hpds.cc/pavement/hpds_node v0.0.0-20230307094826-753c4fe9c877
github.com/go-sql-driver/mysql v1.7.0
github.com/hashicorp/consul/api v1.20.0
github.com/minio/minio-go v6.0.14+incompatible
github.com/spf13/cobra v1.6.1
github.com/spf13/viper v1.15.0
go.uber.org/zap v1.23.0
gopkg.in/yaml.v3 v3.0.1
xorm.io/xorm v1.3.2
)
require (
cloud.google.com/go v0.105.0 // indirect
cloud.google.com/go/compute v1.14.0 // indirect
cloud.google.com/go/compute/metadata v0.2.3 // indirect
cloud.google.com/go/firestore v1.9.0 // indirect
cloud.google.com/go/longrunning v0.3.0 // indirect
git.hpds.cc/Component/mq_coder v0.0.0-20221010064749-174ae7ae3340 // indirect
git.hpds.cc/Component/network v0.0.0-20221012021659-2433c68452d5 // indirect
github.com/armon/go-metrics v0.4.0 // indirect
github.com/coreos/go-semver v0.3.0 // indirect
github.com/coreos/go-systemd/v22 v22.3.2 // indirect
github.com/fatih/color v1.13.0 // indirect
github.com/fsnotify/fsnotify v1.6.0 // indirect
github.com/go-ini/ini v1.67.0 // indirect
github.com/go-task/slim-sprig v0.0.0-20210107165309-348f09dbbbc0 // indirect
github.com/goccy/go-json v0.8.1 // indirect
github.com/gogo/protobuf v1.3.2 // indirect
github.com/golang/groupcache v0.0.0-20210331224755-41bb18bfe9da // indirect
github.com/golang/mock v1.6.0 // indirect
github.com/golang/protobuf v1.5.2 // indirect
github.com/golang/snappy v0.0.4 // indirect
github.com/google/go-cmp v0.5.9 // indirect
github.com/googleapis/enterprise-certificate-proxy v0.2.1 // indirect
github.com/googleapis/gax-go/v2 v2.7.0 // indirect
github.com/hashicorp/go-cleanhttp v0.5.2 // indirect
github.com/hashicorp/go-hclog v1.2.0 // indirect
github.com/hashicorp/go-immutable-radix v1.3.1 // indirect
github.com/hashicorp/go-rootcerts v1.0.2 // indirect
github.com/hashicorp/golang-lru v0.5.4 // indirect
github.com/hashicorp/hcl v1.0.0 // indirect
github.com/hashicorp/serf v0.10.1 // indirect
github.com/inconshreveable/mousetrap v1.0.1 // indirect
github.com/json-iterator/go v1.1.12 // indirect
github.com/lucas-clemente/quic-go v0.29.1 // indirect
github.com/magiconair/properties v1.8.7 // indirect
github.com/marten-seemann/qtls-go1-18 v0.1.2 // indirect
github.com/marten-seemann/qtls-go1-19 v0.1.0 // indirect
github.com/matoous/go-nanoid/v2 v2.0.0 // indirect
github.com/mattn/go-colorable v0.1.12 // indirect
github.com/mattn/go-isatty v0.0.14 // indirect
github.com/mitchellh/go-homedir v1.1.0 // indirect
github.com/mitchellh/mapstructure v1.5.0 // indirect
github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
github.com/modern-go/reflect2 v1.0.2 // indirect
github.com/nxadm/tail v1.4.8 // indirect
github.com/onsi/ginkgo v1.16.4 // indirect
github.com/pelletier/go-toml/v2 v2.0.6 // indirect
github.com/pkg/errors v0.9.1 // indirect
github.com/sagikazarmark/crypt v0.9.0 // indirect
github.com/spf13/afero v1.9.3 // indirect
github.com/spf13/cast v1.5.0 // indirect
github.com/spf13/jwalterweatherman v1.1.0 // indirect
github.com/spf13/pflag v1.0.5 // indirect
github.com/subosito/gotenv v1.4.2 // indirect
github.com/syndtr/goleveldb v1.0.0 // indirect
go.etcd.io/etcd/api/v3 v3.5.6 // indirect
go.etcd.io/etcd/client/pkg/v3 v3.5.6 // indirect
go.etcd.io/etcd/client/v2 v2.305.6 // indirect
go.etcd.io/etcd/client/v3 v3.5.6 // indirect
go.opencensus.io v0.24.0 // indirect
go.uber.org/atomic v1.9.0 // indirect
go.uber.org/multierr v1.8.0 // indirect
golang.org/x/crypto v0.0.0-20220525230936-793ad666bf5e // indirect
golang.org/x/exp v0.0.0-20220722155223-a9213eeb770e // indirect
golang.org/x/mod v0.6.0-dev.0.20220419223038-86c51ed26bb4 // indirect
golang.org/x/net v0.4.0 // indirect
golang.org/x/oauth2 v0.0.0-20221014153046-6fdb5e3db783 // indirect
golang.org/x/sync v0.1.0 // indirect
golang.org/x/sys v0.3.0 // indirect
golang.org/x/text v0.5.0 // indirect
golang.org/x/time v0.1.0 // indirect
golang.org/x/tools v0.1.12 // indirect
golang.org/x/xerrors v0.0.0-20220907171357-04be3eba64a2 // indirect
google.golang.org/api v0.107.0 // indirect
google.golang.org/appengine v1.6.7 // indirect
google.golang.org/genproto v0.0.0-20221227171554-f9683d7f8bef // indirect
google.golang.org/grpc v1.52.0 // indirect
google.golang.org/protobuf v1.28.1 // indirect
gopkg.in/ini.v1 v1.67.0 // indirect
gopkg.in/natefinch/lumberjack.v2 v2.0.0 // indirect
gopkg.in/tomb.v1 v1.0.0-20141024135613-dd632973f1e7 // indirect
xorm.io/builder v0.3.11-0.20220531020008-1bd24a7dc978 // indirect
)
replace git.hpds.cc/pavement/hpds_node => ../hpds_node

View File

@ -0,0 +1,25 @@
package balance
type LbType int
const (
LbRandom LbType = iota
LbRoundRobin
LbWeightRoundRobin
LbConsistentHash
)
func LoadBalanceFactory(lbType LbType) LoadBalance {
switch lbType {
case LbRandom:
return &RandomBalance{}
case LbConsistentHash:
return NewConsistentHashBalance(10, nil)
case LbRoundRobin:
return &RoundRobinBalance{}
case LbWeightRoundRobin:
return &WeightRoundRobinBalance{}
default:
return &RandomBalance{}
}
}

90
internal/balance/hash.go Normal file
View File

@ -0,0 +1,90 @@
package balance
import (
"errors"
"fmt"
"hash/crc32"
"hpds_control_center/model"
"sort"
"strconv"
"sync"
)
// Hash 1 单调性(唯一) 2平衡性 (数据 目标元素均衡) 3分散性(散列)
type Hash func(data []byte) uint32
type UInt32Slice []uint32
func (s UInt32Slice) Len() int {
return len(s)
}
func (s UInt32Slice) Less(i, j int) bool {
return s[i] < s[j]
}
func (s UInt32Slice) Swap(i, j int) {
s[i], s[j] = s[j], s[i]
}
type ConsistentHashBalance struct {
mux sync.RWMutex
hash Hash
replicas int //复制因子
keys UInt32Slice //已排序的节点hash切片
hashMap map[uint32]int64 //节点哈希和key的map, 键是hash值值是节点key
}
func NewConsistentHashBalance(replicas int, fn Hash) *ConsistentHashBalance {
m := &ConsistentHashBalance{
replicas: replicas,
hash: fn,
hashMap: make(map[uint32]int64),
}
if m.hash == nil {
//最多32位保证是一个2^32-1环
m.hash = crc32.ChecksumIEEE
}
return m
}
func (c *ConsistentHashBalance) IsEmpty() bool {
return len(c.keys) == 0
}
// Add 方法用来添加缓存节点参数为节点key比如使用IP
func (c *ConsistentHashBalance) Add(params model.NodeLastStateItem) error {
c.mux.Lock()
defer c.mux.Unlock()
// 结合复制因子计算所有虚拟节点的hash值并存入m.keys中同时在m.hashMap中保存哈希值和key的映射
for i := 0; i < c.replicas; i++ {
hash := c.hash([]byte(strconv.Itoa(i) + fmt.Sprintf("%d", params.NodeId)))
c.keys = append(c.keys, hash)
c.hashMap[hash] = params.NodeId
}
// 对所有虚拟节点的哈希值进行排序,方便之后进行二分查找
sort.Sort(c.keys)
return nil
}
// Get 方法根据给定的对象获取最靠近它的那个节点
func (c *ConsistentHashBalance) Get(key int64) (int64, error) {
if c.IsEmpty() {
return 0, errors.New("node is empty")
}
hash := c.hash([]byte(fmt.Sprintf("%d", key)))
// 通过二分查找获取最优节点,第一个"服务器hash"值大于"数据hash"值的就是最优"服务器节点"
idx := sort.Search(len(c.keys), func(i int) bool { return c.keys[i] >= hash })
// 如果查找结果 大于 服务器节点哈希数组的最大索引,表示此时该对象哈希值位于最后一个节点之后,那么放入第一个节点中
if idx == len(c.keys) {
idx = 0
}
c.mux.RLock()
defer c.mux.RUnlock()
return c.hashMap[c.keys[idx]], nil
}

View File

@ -0,0 +1,8 @@
package balance
import "hpds_control_center/model"
type LoadBalance interface {
Add(model.NodeLastStateItem) error
Get(int64) (int64, error)
}

View File

@ -0,0 +1,32 @@
package balance
import (
"hpds_control_center/model"
"math/rand"
)
// RandomBalance 随机负载均衡
type RandomBalance struct {
curIndex int
rss []int64
}
func (r *RandomBalance) Add(params model.NodeLastStateItem) error {
nodeId := params.NodeId
r.rss = append(r.rss, nodeId)
return nil
}
func (r *RandomBalance) Next() int64 {
if len(r.rss) == 0 {
return 0
}
r.curIndex = rand.Intn(len(r.rss))
return r.rss[r.curIndex]
}
func (r *RandomBalance) Get(int64) (int64, error) {
return r.Next(), nil
}

View File

@ -0,0 +1,35 @@
package balance
import (
"hpds_control_center/model"
)
// RoundRobinBalance 轮询负载均衡
type RoundRobinBalance struct {
curIndex int
rss []int64
}
func (r *RoundRobinBalance) Add(params model.NodeLastStateItem) error {
nodeId := params.NodeId
r.rss = append(r.rss, nodeId)
return nil
}
func (r *RoundRobinBalance) Next() int64 {
if len(r.rss) == 0 {
return 0
}
lens := len(r.rss)
if r.curIndex >= lens {
r.curIndex = 0
}
curNode := r.rss[r.curIndex]
r.curIndex = (r.curIndex + 1) % lens
return curNode
}
func (r *RoundRobinBalance) Get(int64) (int64, error) {
return r.Next(), nil
}

View File

@ -0,0 +1,78 @@
package balance
import (
"hpds_control_center/model"
)
// WeightRoundRobinBalance 加权轮询负载
type WeightRoundRobinBalance struct {
curIndex int
rss []*WeightNode
rsw []int
}
type WeightNode struct {
node model.NodeLastStateItem
Weight int //初始化时对节点约定的权重
currentWeight int //节点临时权重,每轮都会变化
effectiveWeight int //有效权重, 默认与weight相同 , totalWeight = sum(effectiveWeight) //出现故障就-1
}
//1, currentWeight = currentWeight + effectiveWeight
//2, 选中最大的currentWeight节点为选中节点
//3, currentWeight = currentWeight - totalWeight
func (r *WeightRoundRobinBalance) Add(params model.NodeLastStateItem) error {
weightCpu := params.CpuUsed
weightMem := float64(params.MemUsed*100) / float64(params.MemUsed)
weightSwap := float64(params.SwapTotal*100) / float64(params.SwapTotal)
weightTask := 0.0
if len(params.ExecTask) > 0 {
weightTask = 0
} else {
weightTask = 100
}
node := &WeightNode{
node: params,
Weight: int(400 - weightCpu - weightMem - weightSwap - weightTask),
}
node.effectiveWeight = node.Weight
r.rss = append(r.rss, node)
return nil
}
func (r *WeightRoundRobinBalance) Next() int64 {
var best *WeightNode
total := 0
for i := 0; i < len(r.rss); i++ {
w := r.rss[i]
//1 计算所有有效权重
total += w.effectiveWeight
//2 修改当前节点临时权重
w.currentWeight += w.effectiveWeight
//3 有效权重默认与权重相同,通讯异常时-1, 通讯成功+1直到恢复到weight大小
if w.effectiveWeight < w.Weight {
w.effectiveWeight++
}
//4 选中最大临时权重节点
if best == nil || w.currentWeight > best.currentWeight {
best = w
}
}
if best == nil {
return 0
}
//5 变更临时权重为 临时权重-有效权重之和
best.currentWeight -= total
return best.node.NodeId
}
func (r *WeightRoundRobinBalance) Get(int64) (int64, error) {
return r.Next(), nil
}
func (r *WeightRoundRobinBalance) Update() {
}

30
internal/minio/index.go Normal file
View File

@ -0,0 +1,30 @@
package minio
import (
"git.hpds.cc/Component/logging"
"github.com/minio/minio-go"
)
type MinClient struct {
Client *minio.Client
Logger *logging.Logger
}
func NewClient(ak, sak, ep string, useSSL bool, logger *logging.Logger) *MinClient {
client, err := minio.New(ep, ak, sak, useSSL)
if err != nil {
return nil
}
return &MinClient{
Client: client,
Logger: logger,
}
}
func (cli *MinClient) UploadObject(fn, dst, bucket string) error {
_, err := cli.Client.FPutObject(bucket, dst, fn, minio.PutObjectOptions{})
if err != nil {
return err
}
return nil
}

27
main.go Normal file
View File

@ -0,0 +1,27 @@
package main
import (
"fmt"
"github.com/spf13/cobra"
"hpds_control_center/cmd"
"os"
)
var (
rootCmd = &cobra.Command{
Use: "hpds_control_center",
Long: "hpds_control_center is a task control center",
Version: "0.1",
}
)
func init() {
rootCmd.AddCommand(cmd.NewStartCmd())
}
func main() {
if err := rootCmd.Execute(); err != nil {
_, _ = fmt.Fprint(os.Stderr, err.Error())
os.Exit(1)
}
}

42
model/index.go Normal file
View File

@ -0,0 +1,42 @@
package model
import (
"go.uber.org/zap"
"os"
_ "github.com/go-sql-driver/mysql"
"xorm.io/xorm"
"xorm.io/xorm/dialects"
)
var (
DB *xorm.Engine
)
func New(driveName, dsn string, showSql bool) {
DB, _ = NewDbConnection(driveName, dsn)
DB.ShowSQL(showSql)
DB.Dialect().SetQuotePolicy(dialects.QuotePolicyReserved)
err := DB.Sync2(
&IssueModel{},
&Model{},
&Node{},
&NodeState{},
&Task{},
&TaskResult{},
)
if err != nil {
zap.L().Error("同步数据库表结构", zap.Error(err))
os.Exit(1)
}
}
func NewDbConnection(driveName, dsn string) (db *xorm.Engine, err error) {
db, err = xorm.NewEngine(driveName, dsn)
if err != nil {
zap.L().Error("创建数据库连接", zap.Error(err))
os.Exit(-1)
}
db.SetMaxOpenConns(300)
return
}

28
model/model.go Normal file
View File

@ -0,0 +1,28 @@
package model
type Model struct {
ModelId int64 `xorm:"not null pk autoincr INT(11)" json:"modelId"`
ModelName string `xorm:"varchar(200) not null" json:"modelName"`
BizType int `xorm:"int not null default 1" json:"bizType"`
ModelVersion string `xorm:"varchar(50) not null" json:"modelVersion"`
ModelDesc string `xorm:"varchar(200) not null" json:"modelDesc"`
ModelFiles string `xorm:"varchar(200) not null" json:"modelFiles"`
ModelParamsFiles string `xorm:"varchar(200)" json:"modelParamsFiles"`
ModelExecScript string `xorm:"varchar(200)" json:"modelExecScript"`
IsLightWeight bool `xorm:"TINYINT(1) default 0" json:"isLightWeight"`
Status int `xorm:"not null SMALLINT default 0" json:"status"`
CreateAt int64 `xorm:"created" json:"createAt"`
UpdateAt int64 `xorm:"updated" json:"updateAt"`
}
func GetModelById(id int64) *Model {
item := new(Model)
b, err := DB.ID(id).Get(item)
if err != nil {
return nil
}
if !b {
return nil
}
return item
}

11
model/modelIssue.go Normal file
View File

@ -0,0 +1,11 @@
package model
type IssueModel struct {
Id int64 `xorm:"not null pk autoincr INT(11)" json:"id"`
ModelId int64 `xorm:"INT(11) index" json:"modelId"`
NodeId int64 `xorm:"INT(11) index" json:"nodeId"`
Status int `xorm:"not null SMALLINT default 0" json:"status"`
IssueResult string `xorm:"TEXT" json:"issueResult"`
CreateAt int64 `xorm:"created" json:"createAt"`
UpdateAt int64 `xorm:"updated" json:"updateAt"`
}

46
model/node.go Normal file
View File

@ -0,0 +1,46 @@
package model
// Node 节点信息
type Node struct {
NodeId int64 `xorm:"not null pk autoincr INT(11)" json:"nodeId"`
NodeGuid string `xorm:"varchar(100) index" json:"nodeGuid"`
NodeName string `xorm:"varchar(100)" json:"nodeName"`
NodeType int `xorm:"not null SMALLINT default 0" json:"nodeType"`
NodeStatus int `xorm:"not null SMALLINT default 0" json:"nodeStatus"`
Platform string `xorm:"varchar(100)" json:"platform,omitempty"`
PlatformVersion string `xorm:"varchar(100)" json:"platformVersion,omitempty"`
CPU string `xorm:"varchar(1000)" json:"cpu,omitempty"`
MemTotal uint64 `xorm:"BIGINT" json:"memTotal,omitempty"`
DiskTotal uint64 `xorm:"BIGINT" json:"diskTotal,omitempty"`
SwapTotal uint64 `xorm:"BIGINT" json:"swapTotal,omitempty"`
Arch string `xorm:"varchar(1000)" json:"arch,omitempty"`
Virtualization string `xorm:"varchar(1000)" json:"virtualization,omitempty"`
BootTime uint64 `xorm:"BIGINT" json:"bootTime,omitempty"`
IP string `xorm:"varchar(100)" json:"ip"`
IsGpu bool `xorm:"TINYINT(1) default 0" json:"isGpu"`
IsTaskExecute bool `xorm:"TINYINT(1) default 0" json:"isTaskExecute"`
CountryCode string `xorm:"varchar(100)" json:"countryCode,omitempty"`
Version string `xorm:"varchar(100)" json:"version,omitempty"`
CreateAt int64 `xorm:"created" json:"createAt"`
UpdateAt int64 `xorm:"updated" json:"updateAt"`
}
func GetLightWeight(modelId int64) []Node {
list := make([]Node, 0)
err := DB.Table("node").Alias("n").Join("inner", []string{"model_issue", "m"}, "m.node_id = n.node_id").
Where("n.is_task_execute = 1").And("n.is_gpu = 1").And("m.model_id = ?", modelId).Find(&list)
if err != nil {
return nil
}
return list
}
func GetAllNode(modelId int64) []Node {
list := make([]Node, 0)
err := DB.Table("node").Alias("n").Join("inner", []string{"model_issue", "m"}, "m.node_id = n.node_id").
Where("m.model_id = ?", modelId).Find(&list)
if err != nil {
return nil
}
return list
}

60
model/nodeState.go Normal file
View File

@ -0,0 +1,60 @@
package model
// NodeState 节点状态信息
type NodeState struct {
Uptime uint64 `xorm:"BIGINT pk" json:"uptime,omitempty"`
NodeName string `xorm:"varchar(100) pk" json:"nodeName"`
CPU float64 `xorm:"DECIMAL(18,4)" json:"cpu,omitempty"`
MemUsed uint64 `xorm:"BIGINT" json:"memUsed,omitempty"`
SwapUsed uint64 `xorm:"BIGINT" json:"swapUsed,omitempty"`
DiskUsed uint64 `xorm:"BIGINT" json:"diskUsed,omitempty"`
NetInTransfer uint64 `xorm:"BIGINT" json:"netInTransfer,omitempty"`
NetOutTransfer uint64 `xorm:"BIGINT" json:"netOutTransfer,omitempty"`
NetInSpeed uint64 `xorm:"BIGINT" json:"netInSpeed,omitempty"`
NetOutSpeed uint64 `xorm:"BIGINT" json:"netOutSpeed,omitempty"`
Load1 float64 `xorm:"DECIMAL(18,4)" json:"load1,omitempty"`
Load5 float64 `xorm:"DECIMAL(18,4)" json:"load5,omitempty"`
Load15 float64 `xorm:"DECIMAL(18,4)" json:"load15,omitempty"`
TcpConnCount uint64 `xorm:"BIGINT" json:"tcpConnCount,omitempty"`
UdpConnCount uint64 `xorm:"BIGINT" json:"udpConnCount,omitempty"`
ProcessCount uint64 `xorm:"BIGINT" json:"processCount,omitempty"`
}
type NodeLastStateItem struct {
NodeId int64 `json:"nodeId"`
NodeGuid string `json:"nodeGuid"`
NodeName string `json:"nodeName"`
NodeType int `json:"nodeType"`
MemTotal uint64 `json:"memTotal"`
SwapTotal uint64 `json:"swapTotal"`
CpuUsed float64 `json:"cpuUsed"`
MemUsed uint64 `json:"memUsed"`
SwapUsed uint64 `json:"swapUsed"`
Load1 float64 `json:"load1"`
Load5 float64 `json:"load5"`
Load15 float64 `json:"load15"`
ExecTask string `json:"execTask"`
}
func GetNodeState(nodeList []Node) []NodeLastStateItem {
list := make([]NodeLastStateItem, 0)
err := DB.SQL(`select c.node_id,c.node_name,c.node_guid,c.node_type,c.node_type,c.mem_total,
c.swap_total,a.c_p_u cpu_used, a.mem_used,a.swap_used, a.load1, a.load5, a.load15,
d.task_name exec_task from node_state a , (select node_name, max(uptime) uptime from node_state group by node_name) b, node c
left join (select t2.node_id, t2.task_name from task t2, (select node_id, max(start_time) start from task group by node_id) t1
where t2.node_id = t1.node_id and t2.start_time = t1.start and t2.status = 1) d on c.node_id = d.node_id
where a.node_name = b.node_name and a.uptime = b.uptime and a.node_name = c.node_guid and c.node_status > 0 `).Find(&list)
if err != nil {
return nil
}
resList := make([]NodeLastStateItem, 0)
for _, v := range list {
for _, val := range nodeList {
if v.NodeGuid == val.NodeGuid {
resList = append(resList, v)
break
}
}
}
return resList
}

24
model/task.go Normal file
View File

@ -0,0 +1,24 @@
package model
type Task struct {
TaskId int64 `xorm:"not null pk autoincr INT(11)" json:"taskId"`
ModelId int64 `xorm:"INT(11) index" json:"modelId"`
NodeId int64 `xorm:"INT(11) index" json:"nodeId"`
TaskName string `xorm:"VARCHAR(200)" json:"taskName"`
TaskDesc string `xorm:"VARCHAR(500)" json:"taskDesc"`
DatasetArr string `xorm:"TEXT" json:"datasetArr"`
ResultStorage string `xorm:"TEXT" json:"resultStorage"`
AppointmentTime string `xorm:"VARCHAR(30)" json:"appointmentTime"`
StartTime int64 `xorm:"BIGINT" json:"startTime"`
FinishTime int64 `xorm:"BIGINT" json:"finishTime"`
Status int `xorm:"not null SMALLINT default 0" json:"status"` // 1:等待执行; 2:执行中; 3:执行完成; 4:任务分配失败; 5:任务执行失败
CreateAt int64 `xorm:"created" json:"createAt"`
UpdateAt int64 `xorm:"updated" json:"updateAt"`
}
func UpdateTaskExecuteNode(id, nodeId int64) {
item := new(Task)
item.TaskId = id
item.NodeId = nodeId
_, _ = DB.ID(id).Cols("node_id").Update(item)
}

15
model/taskResult.go Normal file
View File

@ -0,0 +1,15 @@
package model
type TaskResult struct {
ResultId int64 `xorm:"not null pk autoincr INT(11)" json:"resultId"`
TaskId int64 `xorm:"INT(11) index" json:"taskId"`
TaskCode string `xorm:"varchar(200)" json:"taskCode"`
ModelId int64 `xorm:"INT(11)" json:"modelId"`
NodeId int64 `xorm:"INT(11)" json:"nodeId"`
StartTime int64 `xorm:"BIGINT" json:"startTime"`
FinishTime int64 `xorm:"BIGINT" json:"finishTime"`
SubDataset string `xorm:"varchar(200)" json:"subDataset"`
DatasetId int64 `xorm:"INT(11)" json:"datasetId"`
SrcPath string `xorm:"varchar(500)" json:"srcPath"`
Result string `xorm:"TEXT" json:"result"`
}

233
mq/index.go Normal file
View File

@ -0,0 +1,233 @@
package mq
import (
"encoding/json"
"fmt"
"git.hpds.cc/Component/logging"
"go.uber.org/zap"
"hpds_control_center/config"
"hpds_control_center/internal/balance"
"hpds_control_center/model"
"os"
"time"
"git.hpds.cc/pavement/hpds_node"
)
var MqList []HpdsMqNode
type HpdsMqNode struct {
MqType uint
Topic string
Node config.HpdsNode
EndPoint interface{}
Logger *logging.Logger
}
func must(logger *logging.Logger, err error) {
if err != nil {
if logger != nil {
logger.With(zap.String("web节点", "错误信息")).Error("启动错误", zap.Error(err))
} else {
_, _ = fmt.Fprint(os.Stderr, err)
}
os.Exit(1)
}
}
func NewMqClient(funcs []config.FuncConfig, node config.HpdsNode, logger *logging.Logger) (mqList []HpdsMqNode, err error) {
mqList = make([]HpdsMqNode, 0)
for _, v := range funcs {
switch v.MqType {
case 2:
sf := hpds_node.NewStreamFunction(
v.Name,
hpds_node.WithMqAddr(fmt.Sprintf("%s:%d", node.Host, node.Port)),
hpds_node.WithObserveDataTags(v.DataTag),
hpds_node.WithCredential(node.Token),
)
err = sf.Connect()
must(logger, err)
nodeInfo := HpdsMqNode{
MqType: 2,
Topic: v.Name,
Node: node,
EndPoint: sf,
}
switch v.Name {
case "task-request":
_ = sf.SetHandler(TaskRequestHandler)
default:
}
mqList = append(mqList, nodeInfo)
default:
ap := hpds_node.NewAccessPoint(
v.Name,
hpds_node.WithMqAddr(fmt.Sprintf("%s:%d", node.Host, node.Port)),
hpds_node.WithCredential(node.Token),
)
err = ap.Connect()
nodeInfo := HpdsMqNode{
MqType: 1,
Topic: v.Name,
Node: node,
EndPoint: ap,
}
must(logger, err)
ap.SetDataTag(v.DataTag)
mqList = append(mqList, nodeInfo)
}
}
return mqList, err
}
func GetMqClient(topic string, mqType uint) *HpdsMqNode {
for _, v := range MqList {
if v.Topic == topic && v.MqType == mqType {
return &v
}
}
return nil
}
func GenerateAndSendData(stream hpds_node.AccessPoint, data []byte) error {
_, err := stream.Write(data)
if err != nil {
return err
}
time.Sleep(1000 * time.Millisecond)
return nil
}
func TaskRequestHandler(data []byte) (byte, []byte) {
fmt.Println("接收数据", string(data))
cmd := new(InstructionReq)
err := json.Unmarshal(data, cmd)
if err != nil {
return 0x0B, []byte(err.Error())
}
switch cmd.Command {
case TaskAdd:
payload := cmd.Payload.(map[string]interface{})
if payload["nodeId"].(float64) == 0 {
//根据业务属性进行分配节点
m := model.GetModelById(int64(payload["modelId"].(float64)))
var nodeList []model.Node
//todo 需要增加模型下发记录
if m.IsLightWeight {
nodeList = model.GetLightWeight(m.ModelId)
} else {
nodeList = model.GetAllNode(m.ModelId)
}
if nodeList != nil {
if len(nodeList) > 1 {
//这里采用加权算法权重采用CPU占用+mem使用+任务执行状态
list := model.GetNodeState(nodeList)
lb := balance.LoadBalanceFactory(balance.LbWeightRoundRobin)
for _, v := range list {
_ = lb.Add(v)
}
nodeId, _ := lb.Get(0)
payload["nodeId"] = nodeId
cmd := &InstructionReq{
Command: TaskExecute,
Payload: payload,
}
pData, _ := json.Marshal(cmd)
cli := GetMqClient("task-execute", 1)
if cli != nil {
_ = GenerateAndSendData(cli.EndPoint.(hpds_node.AccessPoint), pData)
}
model.UpdateTaskExecuteNode(int64(payload["taskId"].(float64)), nodeId)
} else {
payload["nodeId"] = nodeList[0].NodeId
issue := new(model.IssueModel)
h, _ := model.DB.Where("model_id=? and node_id =?", int64(payload["modelId"].(float64)), nodeList[0].NodeId).Get(issue)
if !h {
}
payload["issueResult"] = issue.IssueResult
cmd := &InstructionReq{
Command: TaskExecute,
Payload: payload,
}
pData, _ := json.Marshal(cmd)
cli := GetMqClient("task-execute", 1)
if cli != nil {
_ = GenerateAndSendData(cli.EndPoint.(hpds_node.AccessPoint), pData)
}
model.UpdateTaskExecuteNode(int64(payload["taskId"].(float64)), nodeList[0].NodeId)
}
} else {
}
} else {
cmd := &InstructionReq{
Command: TaskExecute,
Payload: payload,
}
pData, _ := json.Marshal(cmd)
cli := GetMqClient("task-execute", 1)
if cli != nil {
_ = GenerateAndSendData(cli.EndPoint.(hpds_node.AccessPoint), pData)
}
}
case ModelIssue:
payload := cmd.Payload.(map[string]interface{})
cmd := &InstructionReq{
Command: ModelIssueRepeater,
Payload: payload,
}
pData, _ := json.Marshal(cmd)
cli := GetMqClient("task-execute", 1)
if cli != nil {
_ = GenerateAndSendData(cli.EndPoint.(hpds_node.AccessPoint), pData)
}
case ModelIssueResponse:
payload := cmd.Payload.(map[string]interface{})
//查找下发记录
item := new(model.IssueModel)
h, _ := model.DB.Where("model_id = ? and node_id = ?", payload["modelId"].(int64), payload["nodeId"].(int64)).Get(item)
pData, _ := json.Marshal(payload)
if h {
item.Status = 1
item.IssueResult = string(pData)
item.UpdateAt = time.Now().Unix()
_, _ = model.DB.ID(item.Id).AllCols().Update(item)
} else {
item.ModelId = payload["modelId"].(int64)
item.NodeId = payload["nodeId"].(int64)
item.Status = 1
item.IssueResult = string(pData)
item.CreateAt = time.Now().Unix()
item.UpdateAt = time.Now().Unix()
_, _ = model.DB.Insert(item)
}
case TaskResponse:
payload := cmd.Payload.(map[string]interface{})
item := new(model.TaskResult)
item.TaskId = int64(payload["taskId"].(float64))
item.TaskCode = payload["taskCode"].(string)
item.NodeId = int64(payload["nodeId"].(float64))
item.ModelId = int64(payload["modelId"].(float64))
item.StartTime = int64(payload["startTime"].(float64))
item.FinishTime = int64(payload["finishTime"].(float64))
item.SubDataset = payload["subDataset"].(string)
item.DatasetId = int64(payload["datasetArr"].(float64))
item.SrcPath = payload["srcPath"].(string)
item.Result = payload["body"].(string)
_, _ = model.DB.Insert(item)
//fn := payload["fileName"].(string)
//dec := base64.NewDecoder(base64.StdEncoding, strings.NewReader(payload["file"].(string)))
default:
}
return byte(cmd.Command), nil
}

15
mq/instruction.go Normal file
View File

@ -0,0 +1,15 @@
package mq
const (
TaskAdd = iota + 1
ModelIssue
TaskExecute
TaskResponse
ModelIssueRepeater
ModelIssueResponse
)
type InstructionReq struct {
Command int `json:"command"`
Payload interface{} `json:"payload"`
}

View File

@ -0,0 +1,87 @@
package discover
import (
"fmt"
"github.com/hashicorp/consul/api"
)
type ConsulConfig struct {
Client *api.Client `json:"client"` // consul client
ConsulAddress string `json:"consulAddress"` // consul 服务地址:IP+port
ServiceId string `json:"serviceId"` // 服务ID
ServiceName string `json:"serviceName"` // 服务名称
ServiceIP string `json:"serviceIP"` // 服务IP
ServicePort int `json:"servicePort"` // 服务端口
Tags []string `json:"tags"` // 服务标签列表
DeregisterCriticalServiceAfter int `json:"deregisterCriticalServiceAfter"` // 指定与服务关联的检查应在此时间之后注销
Interval int `json:"interval"` // 指定运行此检查的频率
Timeout int `json:"timeout"` // 在脚本、HTTP、TCP 或 gRPC 检查的情况下指定传出连接的超时时间
}
func NewConsulConfig(consulAddress string,
serviceId string,
serviceName string,
serviceIP string,
servicePort int,
tags []string,
deregisterCriticalServiceAfter int,
interval int,
timeout int) (*ConsulConfig, error) {
// 1.consul配置
config := api.DefaultConfig()
config.Address = consulAddress
// 2.client
client, err := api.NewClient(config)
if err != nil {
return nil, err
}
return &ConsulConfig{
Client: client,
ConsulAddress: consulAddress,
ServiceId: serviceId,
ServiceName: serviceName,
ServiceIP: serviceIP,
ServicePort: servicePort,
Tags: tags,
DeregisterCriticalServiceAfter: deregisterCriticalServiceAfter,
Interval: interval,
Timeout: timeout,
}, nil
}
// ServiceRegister 服务注册
func (cf *ConsulConfig) ServiceRegister() (err error) {
// 注册器
reg := &api.AgentServiceRegistration{
ID: cf.ServiceId,
Name: cf.ServiceName,
Address: cf.ServiceIP,
Port: cf.ServicePort,
Tags: cf.Tags,
Check: &api.AgentServiceCheck{
Interval: fmt.Sprintf("%vs", cf.Interval), // 健康检查间隔
HTTP: fmt.Sprintf("http://%v:%v/health", cf.ServiceIP, cf.ServicePort), // HTTP 支持执行健康检查的地址service 会传到 Health.Check 函数中
Timeout: fmt.Sprintf("%vs", cf.Timeout), // 健康检查超时时间
DeregisterCriticalServiceAfter: fmt.Sprintf("%vs", cf.DeregisterCriticalServiceAfter), // 注销时间,相当于过期时间
Notes: "Consul check service health status.",
},
}
// 注册服务
err = cf.Client.Agent().ServiceRegister(reg)
if err != nil {
return err
}
return nil
}
// ServiceDeregister 服务注销
func (cf *ConsulConfig) ServiceDeregister() error {
return cf.Client.Agent().ServiceDeregister(cf.ServiceId)
}
// ServiceDiscover 服务发现
func (cf *ConsulConfig) ServiceDiscover(service string, tags []string, q *api.QueryOptions) ([]*api.CatalogService, *api.QueryMeta, error) {
return cf.Client.Catalog().ServiceMultipleTags(service, tags, q)
}