Question about how to set up multiple different listeners using Service Weaver

php editor Xiaoxin is here to introduce to you how to use Service Weaver to set up multiple different listeners. Service Weaver is a powerful tool that allows developers to create and manage multiple services. By setting different listeners, we can monitor and manage different services. In this article, we will discuss in detail how to use Service Weaver to configure and manage multiple different listeners to better control and optimize our services. Whether you are a beginner or an experienced developer, this article will provide you with detailed guidance and practical tips. Let’s explore together!

Question content

I have been using Service Weaver, which has been released for a while, and was curious how we should use it to set up multiple listeners. My intention is that when we deploy the application, the handler for web and the handler for api run separately (as an example). My code currently looks like this:

package main

import (
    "context"
    "log"
    "sync"

    "github.com/ServiceWeaver/weaver"
    "github.com/labstack/echo/v4"
    "github.com/labstack/echo/v4/middleware"
)

type Server struct {
    weaver.Implements[weaver.Main]
    apiServer weaver.Ref[APIServer]
    webServer weaver.Ref[WebServer]
}

type APIServer interface {
    Serve(context.Context) error
}

type apiServer struct {
    weaver.Implements[APIServer]
    api weaver.Listener
}

func (a apiServer) Serve(ctx context.Context) error {
    logger := a.Logger(ctx)
    e := echo.New()
    e.Listener = a.api

    e.Use(middleware.RequestID())

    e.Use(middleware.RequestLoggerWithConfig(middleware.RequestLoggerConfig{
        LogStatus: true,
        LogURI:    true,
        LogValuesFunc: func(c echo.Context, v middleware.RequestLoggerValues) error {
            logger.Info("Request", "id", v.RequestID, "uri", v.URI, "status", v.Status, "size", v.ResponseSize)
            return nil
        },
    }))

    return e.Start("")
}

type WebServer interface {
    Serve(context.Context) error
}

type webServer struct {
    weaver.Implements[WebServer]
    web weaver.Listener
}

func (w webServer) Serve(ctx context.Context) error {
    logger := w.Logger(ctx)
    e := echo.New()
    e.Listener = w.web

    e.Use(middleware.RequestID())

    e.Use(middleware.RequestLoggerWithConfig(middleware.RequestLoggerConfig{
        LogStatus: true,
        LogURI:    true,
        LogValuesFunc: func(c echo.Context, v middleware.RequestLoggerValues) error {
            logger.Info("Request", "id", v.RequestID, "uri", v.URI, "status", v.Status, "size", v.ResponseSize)
            return nil
        },
    }))

    return e.Start("")
}

func main() {
    if err := weaver.Run(context.Background(), serve); err != nil {
        log.Fatal(err)
    }
}

func serve(ctx context.Context, server *Server) error {
    wg := sync.WaitGroup{}
    wg.Add(2)

    go func() {
        defer wg.Done()
        server.apiServer.Get().Serve(context.Background())
    }()

    go func() {
        defer wg.Done()
        server.webServer.Get().Serve(context.Background())
    }()

    wg.Wait()

    return nil
}

Basically, I just setup two Echo servers, one for each component.

For what it's worth, this is my configuration:

[serviceweaver]
binary = "./platform"

[multi]
listeners.api = { address = "localhost:12345" }
listeners.web = { address = "localhost:54321" }

[single]
listeners.api = { address = "localhost:12345" }
listeners.web = { address = "localhost:54321" }

When I run it in a single process, everything seems to work as expected. I see log entries for network requests indicating that everything is fine. When I run it in deploy mode (i.e., actually doing the multi-process magic), I can only make 1-2 requests before I see a log entry like this, and then the response doesn't work very often:

2023/09/14 21:35:51 http: proxy error: context canceled

It looks like I'm doing something wrong, but it seems like this is a use case that would be supported in some way, so I'm wondering if there's a proper way to solve this problem.

Thanks!

Workaround

tl;dr Service Weaver There is currently no good way to run an HTTP server from a non-primary component. I recommend that you move the two listeners to the Server structure and run the two HTTP servers inside the Serve function. HTTP servers can call methods on other components.

type Server struct {
    weaver.Implements[weaver.Main]
    api weaver.Listener
    web weaver.Listener
}

details

weaver multi The deployer copies each component twice, with each copy running in its own operating system process. Consider what happens when the APIServer component requests a network listener on port 12345. APIServer Two copies of a component cannot listen on port 12345 at the same time; only one operating system process can listen on that port. To solve this problem, two replicas listen on randomly chosen ports, such as 8000 and 9000. The weaver multi deployer then runs an HTTP proxy on port 12345, forwarding requests to ports 8000 and 9000.

In your application, the main Server component is replicated twice, and the Serve function is run twice, once for each replica. Inside Serve, when you call apiServer.Get().Serve(context.Background()), a copy of APIServer is randomly selected for executionServe method. If you're lucky, the two method calls are sent to two different replicas, then everything should run smoothly. However, if both method calls are sent to the same replica, then one replica of APIServer is running an HTTP server and the other is not.

In this case, the proxy forwards half of all requests to the running HTTP server and the other half to the unlistened listener. This will cause you to see proxy errors.

Finally please note that when you run the application using go run or weaver single deploy everything works fine because there is no proxy and the components are not copied.

The above is the detailed content of Question about how to set up multiple different listeners using Service Weaver. For more information, please follow other related articles on the PHP Chinese website!