Logjam client for Go

This package provides integration with the Logjam monitoring tool for Go web-applications.

It buffers all log output for a request and sends the result to a configured logjam device when the request was finished.

Requirements

This package depends on github.com/pebbe/zmq4 which requires ZeroMQ version 4.0.1 or above. Make sure you have it installed on your machine.

E.g. for MacOS:

brew install zmq

How to use it

Install via go get github.com/xing/logjam-agent-go.

Initialize the client

// create a logger
logger := logjam.Logger{
	Logger: log.New(os.Stderr, "", log.StdFlags),
	LogLevel: logjam.ERROR,
}

// create an agent
agent := logjam.NewAgent(&logjam.Options{
	AppName: "MyApp",
	EnvName: "production",
	Logger: logger,
	LogLevel: logjam.INFO,
})

Note that logger and the agent have individual log levels: the one on the logger determines the log level for lines sent to the device it is attached to (os.Stderr in this example), whereas the one on the agent determines which lines are sent to the logjam endpoint.

Use the logjam middleware

r := mux.NewRouter()
r.NotFoundHandler = http.HandlerFunc(logjam.NotFoundHandler)
r.MethodNotAllowedHandler = http.HandlerFunc(logjam.MethodNotAllowedHandler)
...
server := http.Server{
	Handler: agent.NewHandler(r, logjam.MiddlewareOptions{BubblePanics: false})
	...
}

This example uses the Gorilla Mux package but it should also work with other router packages. Don't use the Gorilla syntax: r.Use(agent.NewMiddleware) because middleware added that way is only called for configured routes. So you'll not get 404s tracked in logjam.

You also need to set environment variables to point to the actual logjam broker instance:

export LOGJAM_BROKER=my-logjam-broker.host.name

Shutting down

Make sure to shut down the agent upon program termination in order to properly close the ZeroMQ socket used to send messages to logjam.

agent.Shutdown()

Adapting logjam action names

By default, the logjam middleware fabricates logjam action names from the escaped request path and request method. For example, a GET request to the endpoint "/users/123/friends" will be translated to "Users::Id::Friends#get". If you're not happy with that, you can override the action name in your request handler, as the associated logjam request is available from the request context:

func ShowFriends(w http.ResponseWriter, r *http.Request) {
	request := logjam.GetRequest(r)
	request.ChangeAction(w, "Users#friends")
	fmt.Fprintf(w, "Hello, %q", html.EscapeString(r.URL.Path))
})

If you're using the gorilla mux package, you can configure the desired logjam action name when declaring the route. The following examples uses a Rails inspired naming:

import ("github.com/xing/logjam-agent-go/gorilla")

gorilla.ActionName(router.Path("/users").Methods("GET").HandlerFunc(...), "Users#index")
gorilla.ActionName(router.Path("/users").Methods("POST").HandlerFunc(...), "Users#create")
gorilla.ActionName(router.Path("/users/{user_id}").Methods("GET").HandlerFunc(...), "Users#show")
gorilla.ActionName(router.Path("/users/{user_id}").Methods("PUT", "PATCH").HandlerFunc(...), "Users#update")
gorilla.ActionName(router.Path("/users/{user_id}").Methods("DELETE").HandlerFunc(...), "Users#destroy")
gorilla.ActionName(router.Path("/users/{user_id}/friends").Methods("GET").HandlerFunc(...), "Users#friends")

Make sure to have the route fully configured before calling gorilla.ActionName.

Using the agent for non web requests

The agent's middleware takes care of all the internal plumbing for web requests. If you have some other request handling mechanism, like a message processor for example, you have to manage that yourself.

You will need to create a request, then store the request in a context to be used by the logging mechanism and sending call headers to other services, then perform your business logic and then finally make sure to send the request information to logjam.

// Let's assume that the variable agent points to a logjam.Agent and logger is
// an instance of a logjam.Logger.

func myHandler(ctx context.Context) {
    // create a new request
    request := agent.NewRequest("myHandler#call")
    // create a new context containing the request, so the logjam logger can access it
    ctx := request.NewContext(ctx)
    code := int(200)
    // make sure to send the request at the end
    defer func() {
        // make we send information to logjam in case the app panics
        if r := recover(); r != nil {
            code = 500
            request.Finish(code)
            // optionally reraise the original panic:
            // panic(r)
        }
        request.Finish(code)
    }()
    ...
    // your logic belongs here
    ...
    if sucess {
        logger.Info(ctx, "200 OK")
        return
    }
    code = 500
    logger.Error(ctx, "500 Internal Server Error")
}

Obviously one could abstract this logic into a helper function, but a.t.m. we think this is best left to the application programmer, until we have an agreement on the desired API of such a helper.

Passing call headers to other logjam instrumented services

Logjam can provide caller relationship information between a collection of services, which helps in debugging and understanding your overall system. It can be considered a form of distributed tracing, restricted to HTTP based service relationships. The agent provides a helper function to simplify the task of passing the required information to the called service.

// Call some REST service passing the required logjam headers assuming the variable client
// refers to a http.Client, agent to the logjam agent and ctx refers to the http request
// context of the currently executing request handler which has been augmented by the logjam agent.

req, err := http.NewRequest("GET", "http://example.com", nil)
agent.SetCallHeaders(ctx, req)
resp, err := client.Do(req)

How to contribute?

Please fork the repository and create a pull-request for us.