gRPC負載均衡(自定義負載均衡策略)
阿新 • • 發佈:2020-05-20
### 前言
上篇文章介紹瞭如何實現gRPC負載均衡,但目前官方只提供了`pick_first`和`round_robin`兩種負載均衡策略,輪詢法`round_robin`不能滿足因伺服器配置不同而承擔不同負載量,這篇文章將介紹如何實現自定義負載均衡策略--`加權隨機法`。
`加權隨機法`可以根據伺服器的處理能力而分配不同的權重,從而實現處理能力高的伺服器可承擔更多的請求,處理能力低的伺服器少承擔請求。
### 自定義負載均衡策略
gRPC提供了`V2PickerBuilder`和`V2Picker`介面讓我們實現自己的負載均衡策略。
```go
type V2PickerBuilder interface {
Build(info PickerBuildInfo) balancer.V2Picker
}
```
`V2PickerBuilder`介面:建立V2版本的子連線選擇器。
`Build`方法:返回一個V2選擇器,將用於gRPC選擇子連線。
```go
type V2Picker interface {
Pick(info PickInfo) (PickResult, error)
}
```
`V2Picker `介面:用於gRPC選擇子連線去傳送請求。
`Pick`方法:子連線選擇
問題來了,我們需要把伺服器地址的權重新增進去,但是地址`resolver.Address`並沒有提供權重的屬性。官方給的答覆是:把權重儲存到地址的元資料`metadata`中。
```go
// attributeKey is the type used as the key to store AddrInfo in the Attributes
// field of resolver.Address.
type attributeKey struct{}
// AddrInfo will be stored inside Address metadata in order to use weighted balancer.
type AddrInfo struct {
Weight int
}
// SetAddrInfo returns a copy of addr in which the Attributes field is updated
// with addrInfo.
func SetAddrInfo(addr resolver.Address, addrInfo AddrInfo) resolver.Address {
addr.Attributes = attributes.New()
addr.Attributes = addr.Attributes.WithValues(attributeKey{}, addrInfo)
return addr
}
// GetAddrInfo returns the AddrInfo stored in the Attributes fields of addr.
func GetAddrInfo(addr resolver.Address) AddrInfo {
v := addr.Attributes.Value(attributeKey{})
ai, _ := v.(AddrInfo)
return ai
}
```
定義`AddrInfo`結構體並新增權重`Weight`屬性,`Set`方法把`Weight`儲存到`resolver.Address`中,`Get`方法從`resolver.Address`獲取`Weight`。
解決權重儲存問題後,接下來我們實現加權隨機法負載均衡策略。
首先實現`V2PickerBuilder`介面,返回子連線選擇器。
```go
func (*rrPickerBuilder) Build(info base.PickerBuildInfo) balancer.V2Picker {
grpclog.Infof("weightPicker: newPicker called with info: %v", info)
if len(info.ReadySCs) == 0 {
return base.NewErrPickerV2(balancer.ErrNoSubConnAvailable)
}
var scs []balancer.SubConn
for subConn, addr := range info.ReadySCs {
node := GetAddrInfo(addr.Address)
if node.Weight <= 0 {
node.Weight = minWeight
} else if node.Weight > 5 {
node.Weight = maxWeight
}
for i := 0; i < node.Weight; i++ {
scs = append(scs, subConn)
}
}
return &rrPicker{
subConns: scs,
}
}
```
`加權隨機法`中,我使用空間換時間的方式,把權重轉成地址個數(例如`addr1`的權重是`3`,那麼新增`3`個子連線到切片中;`addr2`權重為`1`,則新增`1`個子連線;選擇子連線時候,按子連線切片長度生成隨機數,以隨機數作為下標就是選中的子連線),避免重複計算權重。考慮到記憶體佔用,權重定義從`1`到`5`權重。
接下來實現子連線的選擇,獲取隨機數,選擇子連線
```go
type rrPicker struct {
subConns []balancer.SubConn
mu sync.Mutex
}
func (p *rrPicker) Pick(balancer.PickInfo) (balancer.PickResult, error) {
p.mu.Lock()
index := rand.Intn(len(p.subConns))
sc := p.subConns[index]
p.mu.Unlock()
return balancer.PickResult{SubConn: sc}, nil
}
```
關鍵程式碼完成後,我們把加權隨機法負載均衡策略命名為`weight`,並註冊到gRPC的負載均衡策略中。
```go
// Name is the name of weight balancer.
const Name = "weight"
// NewBuilder creates a new weight balancer builder.
func newBuilder() balancer.Builder {
return base.NewBalancerBuilderV2(Name, &rrPickerBuilder{}, base.Config{HealthCheck: false})
}
func init() {
balancer.Register(newBuilder())
}
```
完整程式碼[weight.go](https://github.com/Bingjian-Zhu/etcd-example/blob/master/5-etcd-grpclb-balancer/balancer/weight/weight.go)
最後,我們只需要在服務端註冊服務時候附帶權重,然後客戶端在服務發現時把權重`Set`到`resolver.Address`中,最後客戶端把負載論衡策略改成`weight`就完成了。
```go
//SetServiceList 設定服務地址
func (s *ServiceDiscovery) SetServiceList(key, val string) {
s.lock.Lock()
defer s.lock.Unlock()
//獲取服務地址
addr := resolver.Address{Addr: strings.TrimPrefix(key, s.prefix)}
//獲取服務地址權重
nodeWeight, err := strconv.Atoi(val)
if err != nil {
//非數字字元預設權重為1
nodeWeight = 1
}
//把服務地址權重儲存到resolver.Address的元資料中
addr = weight.SetAddrInfo(addr, weight.AddrInfo{Weight: nodeWeight})
s.serverList[key] = addr
s.cc.UpdateState(resolver.State{Addresses: s.getServices()})
log.Println("put key :", key, "wieght:", val)
}
```
客戶端使用`weight`負載均衡策略
```go
func main() {
r := etcdv3.NewServiceDiscovery(EtcdEndpoints)
resolver.Register(r)
// 連線伺服器
conn, err := grpc.Dial(
fmt.Sprintf("%s:///%s", r.Scheme(), SerName),
grpc.WithBalancerName("weight"),
grpc.WithInsecure(),
)
if err != nil {
log.Fatalf("net.Connect err: %v", err)
}
defer conn.Close()
```
執行效果:
執行`服務1`,權重為`1`
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520162934052-74794177.png)
執行`服務2`,權重為`4`
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520162941378-1116335906.png)
執行客戶端
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520163515073-1148862720.png)
檢視前50次請求在`服務1`和`伺服器2`的負載情況。`服務1`分配了`9`次請求,`服務2`分配了`41`次請求,接近權重比值。
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520163753358-1654741743.png)
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520163932810-2034341622.png)
斷開`服務2`,所有請求流向`服務1`
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520164432399-923288256.png)
以權重為`4`,重啟`服務2`,請求以加權隨機法流向兩個伺服器
![](https://img2020.cnblogs.com/blog/1508611/202005/1508611-20200520164648568-1117742551.png)
### 總結
本篇文章以加權隨機法為例,介紹瞭如何實現gRPC自定義負載均衡策略,以滿足我們的需求。
原始碼地址:https://github.com/Bingjian-Zhu/etcd