WrapBackgroundCore background logger is not sending logs to new relic but transaction logger is about go-agent

[yehudamakarov]

@iamemilio working as expected, you are correct!

let's get this in the docs asap! whoever comes next will probably have it way better than me.
thanks a ton for this help.

[iamemilio]

Ah! I know whats happening. Are you waiting for your application to connect to new relic? If not, then its attempting to capture the logs, but failing. We insert our instrumentation into the zapcore object, and anytime a zap process invokes the Write() method, we capture the log in the agent, even if its not synced/flushed yet. But if the agent has not completed its connection handshake, it may drop the logs.

quick fix:

if err := app.WaitForConnection(5 * time.Second); nil != err {
	fmt.Println(err)
}

[yehudamakarov]

@iamemilio this looks very promising, let me try this and get back to you. Can't thank you enough for this! Opening the gates of hope

[yehudamakarov]

my actual code, and this isn't working:

func Logger(app *newrelic.Application) (*zap.Logger, *zapcore.Core, error) {
	core := zapcore.NewCore(zapcore.NewJSONEncoder(zap.NewProductionEncoderConfig()), zapcore.AddSync(os.Stdout), zap.InfoLevel)

	backgroundCore, err := nrzap.WrapBackgroundCore(core, app)
	if err != nil && err != nrzap.ErrNilApp {
		panic(err)
	}

	l, err := zap.New(backgroundCore), nil
	// todo where is this message?
	l.Info("background logs are working")
	return l, &core, err
}

---

but here, the messages arrive

			txn := newrelic.FromContext(ctx)
			if txn == nil {
				txn = app.StartTransaction(req.Spec().Procedure)
			} else {
				txn.SetName(req.Spec().Procedure)
			}

			txnCore, err := nrzap.WrapTransactionCore(*core, txn)
			if err != nil && err != nrzap.ErrNilTxn {
				panic(err)
			}

			txnLogger := zap.New(txnCore)

intermittently I see 1 single message from the original backgroundLogger

[yehudamakarov]

• this may be connected to .Sugar() where anything from a Sugared logger is not in new relic
• log messages in new relic are missing a lot of the json i am sending, i'm looking for any docs on that

[iamemilio]

hmm, let me take a look at this and dig into it a bit. Thanks for reporting!

[iamemilio]

So, I built this reproducer based on your code:

func TestBackgroundLoggerSugared(t *testing.T) {
	app := integrationsupport.NewTestApp(integrationsupport.SampleEverythingReplyFn,
		newrelic.ConfigAppLogDecoratingEnabled(true),
		newrelic.ConfigAppLogForwardingEnabled(true),
	)

	core := zapcore.NewCore(zapcore.NewJSONEncoder(zap.NewProductionEncoderConfig()), zapcore.AddSync(os.Stdout), zap.InfoLevel)

	backgroundCore, err := WrapBackgroundCore(core, app.Application)
	if err != nil && err != ErrNilApp {
		t.Fatal(err)
	}

	logger := zap.New(backgroundCore).Sugar()

	err = errors.New("this is a test error")
	msg := "this is a test error message"

	// for background logging:
	logger.Error(msg, zap.Error(err), zap.String("test-key", "test-val"))
	logger.Sync()

	app.ExpectLogEvents(t, []internal.WantLog{
		{
			Severity:  zap.ErrorLevel.String(),
			Message:   msg,
			Timestamp: internal.MatchAnyUnixMilli,
		},
	})
}

Output:

{"level":"error","ts":1709919339.6399841,"msg":"this is a test error message{error 26 0  this is a test error} {test-key 15 0 test-val <nil>}"}

I do see that in my test function, this log message gets captured by the agent.

got: "this is a test error message{error 26 0  this is a test error} {test-key 15 0 test-val <nil>}"

I am wondering if maybe I am missing something? Could you provide any corrections for the test that you think would cause the error to occur?

[iamemilio]

@yehudamakarov

[yehudamakarov]

I don't know I'll have to check out what a possible difference might be.

There's probably more factors to consider on my end.

[yehudamakarov]

But that first message in my example does not arrive and neither do any from the background logger

[yehudamakarov]

Is it possible it has to do with the logger syncing? Because I even sync it at some point in the startup code (before the tx takes over for requests)

[iamemilio]

hmm, yeah possibly. I don't really have a good idea of how the zap logging buffer gets written, but in the example code I ran I force the buffer to flush by calling Sync(), and that seemed to guarantee capture and writes of those logs.