Function always dies when issuing HTTP(S)-Requests with nodego.OverrideLogger()

[andwun]

I've been having a bad time trying to issue HTTP(S)-Requests from a Cloud Function that uses this package.

Consider this absolute minimal repro:

package main

import (
    "flag"
    "net/http"

    "github.com/GoogleCloudPlatform/cloud-functions-go/nodego"
)

func init() {
    nodego.OverrideLogger()
}

func main() {
    flag.Parse()

    http.HandleFunc(nodego.HTTPTrigger, nodego.WithLoggerFunc(func(w http.ResponseWriter, r *http.Request) {
        http.Get("http://example.com/")
    }))

    nodego.TakeOver()
}

When called, it always dies:

Function execution took 296 ms, finished with status: 'connection error'
Error: function crashed out of request scope Function invocation was interrupted.

Just when writing this, it occured to me to leave out the line nodego.OverrideLogger(). Now the Cloud Function finally does what it is supposed to.

Does anyone have a clue what's wrong here?

[ssttevee]

After a little bit of testing, it seems that this occurs whenever there is any significant waiting time which. However, I was not able to reproduce the same results with several dummy goroutines and the logger goroutine disabled.

I was also still getting the same problem even without the nodego.OverrideLogger() call.

My next suspicion would be the that the instance is actually being killed by a failed report. Some more investigation is needed.

[ssttevee]

I was correct in suspecting that the instance was actually being killed, but it ended up being a problem with error handling that caused it.

[andwun]

Thanks for looking into this problem. I can confirm that the issue is gone when applying your fix.