How do I get the current TraceId?

[fervent valley]

hi,
I am working on an Azure Functions app using Sentry and I want to send the Sentry TraceId to the response so that the end user (the client using my API) can see the TraceId and use it for support. How can I get TraceID? I only see the EventId in the SentrySdk.LastEventId. Thanks!

[sacred flax]

What SDK are you using?

[hardy pendant]

I think something like

var traceHeader = SentrySdk.GetTraceHeader();
var traceId = traceHeader?.TraceId;

should work.

But that's on the .NET Azure Functions app. There you get the extracted ID from the incoming request's headers. And then you want to send it back to the frontend?
Ideally, it'd originate from the frontend anyway. So.. fair What SDK are you using? there?

[fervent valley]

hi @sacred flax and @hardy pendant , thanks for the reply. I am working on a Unity game using Unity SDK with a Azure Functions Isolated Worker backend. I am using 2 different DSNs, which is what was suggeste to me in the chat here when I first set it up, one for the mobile game and one for the backend. When the backend has an error, it generates a traceId which right now is a guid I create, and returns this to the client so the client can do something with it (like report to support for example) right now we are in development so we use it to search the logs to identify root cause of the issue. I would like to avoid having multiple TraceIds, the Sentry one and the application one, and just send the client the TraceId from Sentry and use that in my log searches

Or, if there's a way for my to override sentries TraceId, that would work too, and I could tell Sentry to use my Guid for its traceId

it doesn't seem like GetTraceHeader() is what I need. I need the TraceId that when I search, it pulls up the issue like this:

when I tried using Trace Header, it was a different value

if I use EventId returned from LogException, it will come up in the search based on EventID, but I really don't want to do that, because I want to be able to search for non-failure scenarios this way as well sometimes

[hardy pendant]

Oh I see. Lemme think for a second.

[fervent valley]

hey @hardy pendant , any idea how to get that TraceId I need?

[hardy pendant]

Ah snap! Sorry about that!

So in your system, the Unity game is the origin of those requests, right? So the game does something, it causes an error on your Azure Functions App and you want to connect that request with the error.

[fervent valley]

yeah, basically. So I return a TraceId back to the game client from the server which is a GUID I generate. Actually, I'm using the AzureFunctions_InvocationId which Sentry seems to automatically Tag for me

this way, the user can give me a log or screenshot which contains TraceId for troubleshooting later

[hardy pendant]

Are you using unity web reqests for that?

[fervent valley]

yes

[hardy pendant]

Unfortunately, we don't have proper wrappers setup for that just yet. But you could grab and set the traceID on the Unity side of things and add those to the request.

Sentry then is able to connect those and you can keep using the ID

I've got a sample of pretty much exactly that here

[fervent valley]

awesome thanks, what is the contents of the Baggage thing?

do I need to do something special with these values, or will the Azure Fucntions Sentry middelware autoamtically handle these headers for me?

[hardy pendant]

The baggage contains the sampling decision from the upstream SDK and some dynamic sampling contexts

[hardy pendant]

Yes, it will. That's the idea behind it.

[fervent valley]

nice, I'll playh with these and see what I get

is there any examples of how to best setup Sentry for Unity when you instantiate it and set the config values?

[hardy pendant]

Other than letting it self-initialize? If you need to do any programmatic configuration you can set those build-/runtime configs. Those are scriptable objects that have their config method invoked right before the SDK initializes.
We're splitting it in two (unfortunately) because of mobile, where the native SDKs for Android and iOS do their own thing before the engine even kicks in.

We've got a sample here

[hardy pendant]

Let me know if you run into any problems. More than happy to iterate on the SDK

[fervent valley]

thanks, does the trace header and baggage change on each request?

my game client and game backend should use 2 different DSNs, right?

[fervent valley]

I just tried making 30 API calls in a row in a test, and it seems like it did not generate 30 unique Sentry sentryTraceHeader guid values. It looks like it re-used the values sort of randomly? sometimes 2 consecutive calls used the same one, sometimes it was 4 consecutive calls, sometimes just one call and it changed

Is this what you expected? this seems odd to me, I thought this was for tracing an individual request, so each request would have a unique ID?

[fervent valley]

hey, I think I resolved the issue. I added a transaction at the start and end of each request, and now the SentrySDk.GetTraceHeader() stuff is unique on each request

it's pretty cool! now I can see both backend and game client working together in traces. But what is the gap in between, do you know?

[hardy pendant]

Hmm.. odd, that does not look right. The function is auto-instrumented by the SDK? Or are you creating a transaction yourself?

[fervent valley]

hey @hardy pendant , I'm doing a custom transaction, its because like I said previously, without doing this, Sentry was re-using the trace headers for multiple requests, so I have to manually create a transaction on each request. I got this idea from the sample you showed so I'm guessing thats "by design"

I moved around some tracing code, does this look right now?

[hardy pendant]

It does. I'm just curious to know what's up with those huge delays before and after the function does its thing.

[fervent valley]

oh, I thought this is just showing the full trace time? for example, the client took 336MS to refresh the catalog, and 9MS of that was spent on the server?

the stuff before and after that I assume is the game client code doing things like making the network call, parsing, etc

[hardy pendant]

Oh yeah, there's just that 200ms gap between you starting the transaction and the function starting to do its thing.

Fair enough, actually

But the relationship now looks like it's supposed to

[fervent valley]

awesome, thank you for confirming

one thing I noticed is the UserId is some random guid on the client. I am not setting it explicitly

on the server, its working as expected as I am setting it explicitly

do you know where that random UserId comes from on the client?

[hardy pendant]

It's an installation ID. Depending on the platform the SDK is either creating that GUID itself or it's relying on UnityAnalytics to provide one.

[fervent valley]

wow, the hierarchy and screenshot feature of Sentry is freaking incredible

the one thing I'm not seeing though is the server side of failures. For example, client sends a web request to server, includes the baggage/traceID, and server logs an error, client receives the error and it also logs an error. I am now looking at the clients error, and do not see any links to the server

of course, since I actually print a traceId to the message (I have my server send my own custom TraceID to client which I log on both sides), I can find it by searching the server logs with that TraceId, but I'm surprised that I dont just see the automatic connection in the error log for the client

[hardy pendant]

There should be a "View Full Trace" at the top left of the issue

[fervent valley]

thanks @hardy pendant , I didn't see the server side in the full trace. I'll try to send an example

I am also getting some warnings, so maybe I am doing something wrong

for example, I get this : "Sentry: (Warning) Push scope called in global mode, returning. "

[fervent valley]

I also get this warning, I posted in a different thread:

UnityEngine.DebugLogHandler:LogFormat (UnityEngine.LogType,UnityEngine.Object,string,object[])
Sentry.Unity.Integrations.UnityLogHandlerIntegration:LogFormat (UnityEngine.LogType,UnityEngine.Object,string,object[]) (at /sentry-unity/src/Sentry.Unity/Integrations/UnityLogHandlerIntegration.cs:99)
UnityEngine.Logger:Log (UnityEngine.LogType,string,object)
Sentry.Unity.UnityLogger:Log (Sentry.SentryLevel,string,System.Exception,object[]) (at /sentry-unity/src/Sentry.Unity/UnityLogger.cs:33)
Sentry.Extensibility.DiagnosticLoggerExtensions:LogIfEnabled (Sentry.Extensibility.IDiagnosticLogger,Sentry.SentryLevel,System.Exception,string) (at /sentry-unity/src/sentry-dotnet/src/Sentry/Extensibility/DiagnosticLoggerExtensions.cs:428)
Sentry.Extensibility.DiagnosticLoggerExtensions:LogWarning (Sentry.SentryOptions,string) (at /sentry-unity/src/sentry-dotnet/src/Sentry/Extensibility/DiagnosticLoggerExtensions.cs:178)
Sentry.GlobalSessionManager:ResumeSession () (at /sentry-unity/src/sentry-dotnet/src/Sentry/GlobalSessionManager.cs:384)
Sentry.Internal.Hub:ResumeSession () (at /sentry-unity/src/sentry-dotnet/src/Sentry/Internal/Hub.cs:320)
Sentry.Unity.Integrations.SessionIntegration/<>c__DisplayClass2_0:<Register>b__0 () (at /sentry-unity/src/Sentry.Unity/Integrations/SessionIntegration.cs:27)
Sentry.Unity.SentryMonoBehaviour:UpdatePauseStatus (bool) (at /sentry-unity/src/Sentry.Unity/SentryMonoBehaviour.cs:86)
Sentry.Unity.SentryMonoBehaviour:OnApplicationFocus (bool) (at /sentry-unity/src/Sentry.Unity/SentryMonoBehaviour.cs:99)
UnityEditor.HostView:OnFocus ()```

[fervent valley]

@hardy pendant it seems like I had to register things twice to get everything working in my azure functions startup

no idea why, but now profiling and ful trace is working

here's what my program.cs looks like now. I had to add the builder.UseSentry() with all the options, and then the plumbing seems to work now fully. PReviously I was using logging.AddSentry only. I am doing the logging.AddSentry since I am using the MS ILogging integration which I thought made that part required

[hardy pendant]

Hey @fervent valley, sorry, I just got back, let me catch up here

[hardy pendant]

It looks like the Unity SDK is missing some special session handling for errors. Thanks for raising this.