High Trace Duration in Distributed Tracing When App Goes to Background

[ngcvm]

Description
Not sure if it's an issue. I'm using newrelic-react-native-agent as a New Relic agent in my React Native project. When viewing the distributed tracing for my mobile app in the New Relic dashboard (Production), I notice that some tracing requests have unusually high trace duration times.

After debugging locally, I can confirm that my HTTP requests do not take that long, and the test distributed tracing logs reflect the correct durations. However, I can reproduce the issue by opening the app and suddenly putting it into the background.

In the Timeline view of the high trace duration events, I see a blank time gap between the start span and end span, which seems to correspond to the time the app was in the background.

Context

• React Native version: 0.75.3
• newrelic-react-native-agent version: 1.5.5
• Platform: iOS
• Background fetch: Enabled

Expected Behavior
The trace duration time should reflect the actual HTTP request time, even if the app transitions to the background during the request.

Observed Behavior

• Trace duration times are significantly inflated when the app transitions to the background during an HTTP request.
• In the Timeline view, there is a blank time gap between the start span and end span, which seems to correspond to the time the app was in the background.

Questions

1. Does the trace duration time also include UI interaction or other app lifecycle events?
2. If I want to track only the HTTP request time for mobile, should I use custom events instead of relying on noticeHttpTransaction?

Steps to Reproduce

1. Open the app and initiate an HTTP request.
2. Immediately put the app into the background while the request is in progress.
3. Bring the app back to the foreground and observe the distributed tracing logs in the New Relic dashboard.

Additional Notes

• The issue seems to occur because the app transitions to the background, and the trace duration includes the time the app was paused.
• I have tried to disabled these configurations: interactionTracingEnabled, webViewInstrumentation

[mbruin-NR]

I suspect that you're correct, if you background the app the thread that the web request is on gets paused before it finishes and then it would records the total time when the app foregrounds and the thread continues.
What priority is the thread that you are calling the web request? A higher priority may keep the thread active for longer on backgrounding.
My example app always finishes the web requests and records an accurate time. Could you provide a simple example app that reproduces this issue?

[ngcvm]

I am currently utilizing axios: 1.8.4 and @tanstack/react-query: 4.36.1 as my HTTP client setup. My configuration is set to default, and I have not explicitly defined any thread priorities at this time.

1. Could you elaborate on how thread priority functions within React Native?
2. I have observed that HTTP requests continue to complete successfully even after the application is sent to the background. This is confirmed by logging within the response interceptor, and I have enabled Background Fetch for the iOS app. I suspect the issue may be that the agent necessitates UI interaction between the root span and the final span to mark an event as completed.
3. Unfortunately, I cannot provide an example application at the moment, but I will attempt to prepare one, although this is not guaranteed.
4. Are there any alternative solutions or workarounds to address this problem using other features, such as custom events or OpenTelemetry?

Thank you for your assistance.

[mbruin-NR]

Can you collect some Charles logs so that we can see the data being sent to the server and confirm it's look right? If the web requests are completing it is unlikely that the time tracker would be off because the agent swizzles those network requests.
You mentioned using noticeHttpTransaction can you confirm the values for start and end time are correct? Also that shouldn't be needed because the agent should automatically track network requests. Doing both might confuse the backend. You can try to disable the automatic capture by disabling networkRequestEnabled, networkErrorRequestEnabled, and httpResponseBodyCaptureEnabled or remove the noticeHttpTransaction.

[ngcvm]

Hi @mbruin-NR, I hope these logs are what you mean.
I have removed the calls to noticeHttpTransaction-> logs.zip
I have disabled networkRequestEnabled and networkErrorRequestEnabled and use noticeHttpTransaction => no logs are reported on NewRelic => disable-networkRequestEnabled.chlz.zip

logs.zip

disable-networkRequestEnabled.chlz.zip