想了解分发任务的通信架构如何实现

[fqzhao-win]

ven-worker-sink的通信架构在clip-as-service中仍然保留嘛，在哪个部分的代码，想要具体学习一下这个部分

[jemmyshin]

这块具体的逻辑在jina内部，clip-as-service是对jina内部逻辑的封装，想要了解细节的话你可以具体看一下jina flow的文档：https://docs.jina.ai/concepts/flow/

[fqzhao-win]

谢谢您，我刚刚看了flow的文档，想问一下您flow支持添加http协议的外部程序嘛

[fqzhao-win]

from jina import Flow,Document

replica_hosts, replica_ports = ['localhost', '91.198.174.192'], ['12345', '12346']

Flow().add(host=replica_hosts, port=replica_ports, external=True)

alternative syntax

def main(): f = Flow().add(host=[' http://192.168.0.243:8036', ' http://192.168.0.243:8038'], external=True) print(f.is_flow_ready())

if name == 'main': main()

[fqzhao-win]

这里不知道为什么打印出的是Flase

[fqzhao-win]

ERROR GRPCClient@3104 RpcError: failed to connect to all [01/04/23 15:29:23] addresses

[numb3r3]

basically, we only support grpc-based external service. Thus, the right approach is to start external service ahead:

# the flow uses default `grpc` protocol
with Flow(port='12345').add(....) as external_flow:
    # specify the external service host and port
    with Flow().add(host='0.0.0.0', port=external_flow.port, external=True) as f:
        f.post(...)

[numb3r3]

The communication among gateway, and (external-)executors are all grpc-based protocol.

[fqzhao-win]

ok,thank you very much

[fqzhao-win]

不好意思再麻烦一下大家，如果我现在有一个flask服务，我在gpu上面复制了多个相同程序，分别占用不同的端口，想用ven-work-sink通信架构来解决并发问题，我可以用flow来实现嘛，把flask的内部程序做成Executor

[numb3r3]

是的，完全可以用executor来解决并发扩展问题。

[fqzhao-win]

今天看了一天文档，但是还是没有思路，再麻烦一下你们，就是我有一个服务，这个服务的参数是一个字符串列表，想用flow实现负载均衡，jina可以实现自动对列表切分，然后把结果拼接起再返回吗，我只需要给一个切分的值，不需要做其他设计可以实现吗，头都要炸了

[numb3r3]

closed by https://github.com/jina-ai/jina/issues/5575