聊聊dubbo-go的forkingCluster

序

本文主要研究一下dubbo-go的forkingCluster

forkingCluster

dubbo-go-v1.4.2/cluster/cluster_impl/forking_cluster.go

代码语言：javascript复制

type forkingCluster struct{}

const forking = "forking"

func init() {
    extension.SetCluster(forking, NewForkingCluster)
}

// NewForkingCluster ...
func NewForkingCluster() cluster.Cluster {
    return &forkingCluster{}
}

func (cluster *forkingCluster) Join(directory cluster.Directory) protocol.Invoker {
    return newForkingClusterInvoker(directory)
}

• forkingCluster的Join方法执行newForkingClusterInvoker

newForkingClusterInvoker

dubbo-go-v1.4.2/cluster/cluster_impl/forking_cluster_invoker.go

代码语言：javascript复制

type forkingClusterInvoker struct {
    baseClusterInvoker
}

func newForkingClusterInvoker(directory cluster.Directory) protocol.Invoker {
    return &forkingClusterInvoker{
        baseClusterInvoker: newBaseClusterInvoker(directory),
    }
}

• newForkingClusterInvoker创建了forkingClusterInvoker

Invoke

dubbo-go-v1.4.2/cluster/cluster_impl/forking_cluster_invoker.go

代码语言：javascript复制

// Invoke ...
func (invoker *forkingClusterInvoker) Invoke(ctx context.Context, invocation protocol.Invocation) protocol.Result {
    err := invoker.checkWhetherDestroyed()
    if err != nil {
        return &protocol.RPCResult{Err: err}
    }

    invokers := invoker.directory.List(invocation)
    err = invoker.checkInvokers(invokers, invocation)
    if err != nil {
        return &protocol.RPCResult{Err: err}
    }

    var selected []protocol.Invoker
    forks := int(invoker.GetUrl().GetParamInt(constant.FORKS_KEY, constant.DEFAULT_FORKS))
    timeouts := invoker.GetUrl().GetParamInt(constant.TIMEOUT_KEY, constant.DEFAULT_TIMEOUT)
    if forks < 0 || forks > len(invokers) {
        selected = invokers
    } else {
        selected = make([]protocol.Invoker, 0)
        loadbalance := getLoadBalance(invokers[0], invocation)
        for i := 0; i < forks; i   {
            ivk := invoker.doSelect(loadbalance, invocation, invokers, selected)
            if ivk != nil {
                selected = append(selected, ivk)
            }
        }
    }

    resultQ := queue.New(1)
    for _, ivk := range selected {
        go func(k protocol.Invoker) {
            result := k.Invoke(ctx, invocation)
            err := resultQ.Put(result)
            if err != nil {
                logger.Errorf("resultQ put failed with exception: %v.n", err)
            }
        }(ivk)
    }

    rsps, err := resultQ.Poll(1, time.Millisecond*time.Duration(timeouts))
    if err != nil {
        return &protocol.RPCResult{
            Err: fmt.Errorf("failed to forking invoke provider %v, " 
                "but no luck to perform the invocation. Last error is: %v", selected, err),
        }
    }
    if len(rsps) == 0 {
        return &protocol.RPCResult{Err: fmt.Errorf("failed to forking invoke provider %v, but no resp", selected)}
    }

    result, ok := rsps[0].(protocol.Result)
    if !ok {
        return &protocol.RPCResult{Err: fmt.Errorf("failed to forking invoke provider %v, but not legal resp", selected)}
    }

    return result
}

• Invoke方法先通过invoker.directory.List(invocation)获取invokers，之后从invoker.GetUrl()获取forks及timeouts参数，然后循环forks次通过invoker.doSelect(loadbalance, invocation, invokers, selected)选出selected的invokers；之后遍历selected异步执行其Invoke方法，并将结果放到resultQ中；最后通过resultQ.Poll(1, time.Millisecond*time.Duration(timeouts))拉取最先返回的结果返回

小结

forkingCluster的Join方法执行newForkingClusterInvoker；其Invoke方法循环forks次通过invoker.doSelect(loadbalance, invocation, invokers, selected)选出selected的invokers；之后遍历selected异步执行其Invoke方法，并将结果放到resultQ中；最后通过resultQ.Poll(1, time.Millisecond*time.Duration(timeouts))拉取最先返回的结果返回

doc

• forking_cluster

unix dubbo go

0 人点赞