Performance Monitoring troubleshooting and FAQ

How to view log messages for performance events

1. Enable debug logging for Performance Monitoring at build time by adding a <meta-data> element to your app's AndroidManifest.xml file, like so:

   <application>
       <meta-data
         android:name="firebase_performance_logcat_enabled"
         android:value="true" />
   </application>
   

2. Check your log messages for any error messages.

3. Performance Monitoring tags its log messages with FirebasePerformance. Using logcat filtering, you can specifically view duration trace and HTTP/S network request logging by running the following command:

   adb logcat -s FirebasePerformance

4. Check for the following types of logs which indicate that Performance Monitoring is logging performance events:

   • Logging trace metric: TRACE_NAME, FIREBASE_PERFORMANCE_CONSOLE_URL
   • Logging network request trace: URL

5. Click on the URL to view your data in the Firebase console. It may take a few moments for the data to update in the dashboard.

If your app isn't logging performance events, review the troubleshooting tips.

Added the SDK to app, but console still says to add SDK

Firebase can detect if you've successfully added the Performance Monitoring SDK to your app when it receives event information (like app interactions) from your app. Usually within 10 minutes of starting your app, the Performance dashboard of the Firebase console displays an "SDK detected" message. Then, within 30 minutes, the dashboard displays the initial processed data.

If it's been more than 10 minutes since you added the latest version of SDK to your app, and you're still not seeing any change, check your log messages to make sure that Performance Monitoring is logging events. Try the appropriate troubleshooting steps as described below to troubleshoot a delayed SDK detection message.

Console says the SDK is detected, but no data is displayed

Performance Monitoring processes performance event data before displaying it in the Performance dashboard.

If it's been more than 24 hours since the "SDK detected" message appeared, and you're still not seeing data, then check the Firebase Status Dashboard in case there is a known outage. If there is no outage, contact Firebase Support.

App is not logging performance events

If you're not seeing log messages for performance events, try the following troubleshooting steps:

1. Check the setup of the Performance Monitoring Gradle plugin, as follows:

   1. Make sure that you added the plugin correctly. Specifically, check the following:

      • You added the plugin (apply plugin: 'com.google.firebase.firebase-perf') in your module (app-level) build.gradle file.
      • You included the classpath dependency for the plugin (classpath 'com.google.firebase:perf-plugin:1.4.2') in your project-level build.gradle file.

   2. Make sure that the plugin is not disabled through either of the following flags:

      • instrumentationEnabled in your module (app-level) build.gradle file
      • firebasePerformanceInstrumentationEnabled in your gradle.properties file

2. Check that the Performance Monitoring SDK is not disabled through either of the following flags in your AndroidManifest.xml file:

   • firebase_performance_collection_enabled
   • firebase_performance_collection_deactivated

3. Make sure that Performance Monitoring is not disabled at runtime.

4. If you can't find anything that's disabled in your app, contact Firebase Support.

Performance dashboard is missing screen trace data

If you're missing data for screen rendering traces, try the following troubleshooting steps:

1. Make sure that you're using the latest version of the Android SDK (v21.0.2). Screen rendering traces are only available with v15.2.0 or later.

2. Make sure that you haven't manually disabled Hardware Acceleration for a screen.

3. Make sure that you're not using DexGuard or Jack. Performance Monitoring is incompatible with these toolchains.

   • DexGuard disables automatic collection of app start, app-in-foreground, and app-in-background traces. However, any custom code traces should behave normally if your app uses DexGuard.

   • Jack is deprecated and generally should not be used in your app.

Performance dashboard is missing custom trace data

Are you seeing performance data for automatically collected traces but not for custom code traces? Try the following troubleshooting steps:

1. If you instrumented custom code traces via the Trace API, check the setup of the traces, especially the following:

   • Names for custom code traces and custom metrics must meet the following requirements: no leading or trailing whitespace, no leading underscore (_) character, and max length is 32 characters.
   • All traces must be started and stopped. Any trace that is not started, not stopped, or stopped before started will not be logged.

2. If you instrumented custom code traces via @AddTrace notation, check the setup of the Performance Monitoring Gradle plugin:

   1. Make sure that you added the plugin correctly. Specifically, check the following:

      • You added the plugin (apply plugin: 'com.google.firebase.firebase-perf') in your module (app-level) build.gradle file.
      • You included the classpath dependency for the plugin (classpath 'com.google.firebase:perf-plugin:1.4.2') in your project-level build.gradle file.

   2. Make sure that the plugin is not disabled through either of the following flags:

      • instrumentationEnabled in your module (app-level) build.gradle file
      • firebasePerformanceInstrumentationEnabled in your gradle.properties file

3. Check your log messages to make sure that Performance Monitoring is logging expected custom code traces.

4. If Performance Monitoring is logging events, but no data displays after 24 hours, contact Firebase Support.

Performance dashboard is missing network request data

If you're missing network request data, try the following troubleshooting steps:

1. For Android apps, the Performance Monitoring Gradle plugin enables instrumentation that provides automatic monitoring of HTTP/S network requests. Check the following:

   1. Make sure that you added the plugin correctly. Specifically, check the following:

      • You added the plugin (apply plugin: 'com.google.firebase.firebase-perf') in your module (app-level) build.gradle file.
      • You included the classpath dependency for the plugin (classpath 'com.google.firebase:perf-plugin:1.4.2') in your project-level build.gradle file.

   2. Make sure that the plugin is not disabled through either of the following flags:

      • instrumentationEnabled in your module (app-level) build.gradle file
      • firebasePerformanceInstrumentationEnabled in your gradle.properties file

2. Check for network library incompatibility. Performance Monitoring automatically collects metrics for network requests that use the following networking libraries: OkHttp 3.x.x, Java's URLConnection, and Apache HttpClient.

   Note that you can add custom monitoring for network requests.

3. Be aware of the following:

   • Depending on the behavior of your code and networking libraries used by your code, Performance Monitoring might only report on network requests that are completed. This means that HTTP/S connections that are left open might not be reported.

   • Performance Monitoring is not compatible with DexGuard and Jack.

     • DexGuard disables monitoring of HTTP/S network requests.
     • Jack is deprecated and generally should not be used in your app.

   • Performance Monitoring does not report on network requests with invalid Content-Type headers. However, network requests without the Content-Type headers will still be accepted.

Network request data are not aggregating as expected

Learn more about how Performance Monitoring aggregates network request data under URL patterns.

You can also try out custom URL patterns!

What happened to Top Issues in the Performance card on Project home?

We replaced Top Issues with Recent Alerts as a follow-up to our recent introduction of alerts, which automatically notify you when the thresholds you set are crossed. Issues are now deprecated and replaced by alerts.

The apps selector at the top of the Performance card filters the alert entries under Recent Alerts. Only the three most recent alerts for the app(s) selected are displayed.

To learn more about alerts, see Set up alerts for performance issues.

What happened to the ability to set thresholds for issues in the console?

Performance Monitoring supports alerts for metrics that exceed defined thresholds. To avoid confusion with these configurable thresholds for performance metrics, we removed the ability to configure thresholds for issues.

What happened to the Details and Metrics information in the Firebase console?

We replaced the Details and Metrics pages with a newly redesigned, centralized user interface (UI) to improve how you troubleshoot issues. This new troubleshooting UI offers the same core functionality that Details and Metrics offered. To learn more about troubleshooting, see View more data for a specific trace.

Why is the number of samples not what I expect?

Performance Monitoring collects performance data from your app’s user devices. If your application has many users or if the app generates a large amount of performance activity, Performance Monitoring might limit data collection to a subset of devices to reduce the number of processed events. These limits are high enough so that, even with fewer events, the metric values are still representative of your user's app experience.

To manage the volume of data that we collect, Performance Monitoring uses the following sampling options:

• On-device rate limiting: To prevent a device from sending sudden bursts of traces, we limit the number of code and network request traces sent from a device to 300 events every 10 mins. This approach protects the device from looped instrumentations that can send large amounts of performance data, and it prevents a single device from skewing the performance measurements.

• Dynamic sampling: Performance Monitoring collects a limited number of code traces and network request traces per app daily across all app users. A dynamic sampling rate is fetched on devices (using Firebase Remote Config) to determine whether a random device should capture and send traces. A device that is not selected for sampling does not send any events. The dynamic sampling rate is app-specific and adjusts to ensure that the overall volume of collected data remains below the limit.

  Projects that enabled BigQuery integration receive a higher limit for the number of network request traces.

  User sessions send additional, detailed data from a user's device, requiring more resources to capture and send the data. To minimize the impact of user sessions, Performance Monitoring might also restrict the number of sessions.

• Server-side rate limiting: To ensure that apps don't exceed the sampling limit, Performance Monitoring might use server-side sampling to drop some events received from devices. Although this type of limiting doesn't change the effectiveness of our metrics, it may cause minor pattern shifts, including the following:

  • The number of traces can differ from the number of times that a piece of code was executed.
  • Traces that are closely coupled in code may each have a different number of samples.

What happened to the Issues tab in the console?

We replaced the Issues tab with the introduction of Alerts, which automatically notifies you when the thresholds you set are exceeded. You no longer need to manually check the Firebase console to determine the status of a threshold. To learn about Alerts, see Set up alerts for performance issues.

What happened to the On Device and Network tabs in the console? How do I view the traces that were on those pages?

We've redesigned the Performance Monitoring section of the Firebase console so that the Dashboard tab displays your key metrics and all your traces in one space. As part of the redesign, we removed the On device and Network pages.

The traces table at the bottom of the Dashboard tab has all the same information that the On device and Network tabs displayed, but with some added features, including the ability to sort your traces by the percentage change for a specific metric. To view all the metrics and data for a specific trace, click the trace name in the traces table.

View your traces in the following subtabs of the traces table:

• Network request traces (both out-of-the-box and custom) — Network requests subtab
• Custom code traces — Custom traces subtab
• App start, app-in-foreground, app-in-background traces — Custom traces subtab
• Screen rendering traces — Screen rendering subtab
• Page load traces — Page load subtab

For details about the traces table and viewing metrics and data, visit the console overview page (iOS+ | Android | Web).

Why is the number of slow and frozen frames not what I expected?

Slow rendering frames and frozen frames are calculated with an assumed device refresh rate of 60Hz. If a device refresh rate is lower than 60Hz, each frame will have a slower rendering time because fewer frames are rendered per second. Slower rendering times can cause more slow or frozen frames to be reported because more frames will be rendered slower or will freeze. However, if a device refresh rate is higher than 60Hz, each frame will have a faster rendering time. This can cause fewer slow or frozen frames to be reported. This is a current limitation in the Performance Monitoring SDK.

Why am I unable to see fragment traces?

To see the performance of fragments in addition to app activity, make sure that your app is using Performance Monitoring Android SDK version 20.1.0 or above. To learn more, see Add Performance Monitoring to your app.

How do I understand which traces are related to fragments and activities?

Each of the fragment and activity traces is based on its class name as defined in your application. Each of the screen traces contains the st prefix followed by the name of the class. On the Firebase console, the prefix is removed. To learn more, see Learn about screen rendering performance data (Apple & Android apps) .

Why am I seeing fewer fragment traces than other traces?

Performance Monitoring conducts event sampling across all events collected on a device. This approach lets us collect the minimum events needed from user devices to provide performance metrics.

How am I notified when there is an issue with the rendering performance of my application?

Performance Monitoring lets you set up alerts for the metrics you care about. For generated screen rendering traces, you can set up alerts to notify you when the slow and frozen frames percentage exceeds a threshold that you set.

My build times are high after enabling the Performance Monitoring Gradle plugin. How can I improve this?

Performance Monitoring for Android uses bytecode instrumentation to provide some out-of-the-box features like monitoring HTTP/S network requests. As a part of compilation, the process requires iteration through all the classes of your app (including dependencies) to instrument the code that is crucial in measuring the network request performance of your application.

Here are some key contributors to an increase in build time:

• Number of classes or files
• Size of each of those classes (lines of code)
• Your machine configuration
• Initial build versus a subsequent build (subsequent builds are usually faster than the initial build)

To optimize your build time, consider modularizing your code.

Starting with v1.3.3 of the Performance Monitoring plugin, we've concentrated on making considerable improvements in the incremental build processing and caching of library inputs. To receive the most recent build time improvements, make sure to use the latest version of the plugin (v1.4.2).

Note that you can disable the Performance Monitoring plugin for your debug builds locally if you want to avoid long build times. However, this approach is not recommended for production builds, as it could result in missed performance measurements for the network requests in your app.

What do I do if I get build errors due to incompatible libraries with the Performance Monitoring Gradle plugin?

Performance Monitoring for Android uses bytecode instrumentation to provide some out-of-the-box features like monitoring HTTP/S network requests. As a part of compilation, the process requires iteration through all the classes of your app (including dependencies) to instrument the code that is crucial in measuring the network request performance of your application.

If you get build errors like JSR/RET are not supported with computeFrames option or similar errors after integrating with the Performance Monitoring plugin, this might be because you also have a dependency on a library that's incompatible with the Performance Monitoring Gradle plugin.

To get around this, you can exclude incompatible classes/libraries from being instrumented by following these steps:

1. Update to the latest version of Performance Monitoring Gradle plugin (minimum v1.4.0).
2. Update your Android Gradle plugin version to v7.2.0 or newer.
3. Add the following flag to your module (app-level) build.gradle file to exclude the incompatible classes/libraries from being instrumented:

   android {
     // ...
     androidComponents {
       onVariants(selector().all(), {
           instrumentation.excludes.add("example.incompatible.library")
       })
     }
   }
   

   To learn more about the exclude property of Android Gradle plugin's Instrumentation API, see Instrumentation.

Please file a Github issue when you encounter build errors due to incompatible libraries so that they can also be excluded from being instrumented in the Performance Monitoring plugin.

My Performance Monitoring data is taking longer than expected to be exported to BigQuery. Is it not real time?

If you have enabled the BigQuery integration for Firebase Performance Monitoring, your data will be exported to BigQuery 12 to 24 hours after the end of the day (Pacific Time).

For example, the data for April 19th will be available in BigQuery on April 20th between 12:00pm and midnight (all dates and times are Pacific Time).

What does "near real-time" performance data mean?

Firebase Performance Monitoring processes collected performance data as it comes in, which results in near real-time data display in the Firebase console. Processed data displays in the console within a few minutes of its collection, hence the term "near real-time".

To take advantage of near real-time data processing, make sure your app uses a real-time compatible SDK version.

How do I get near real-time performance data for my app?

To take advantage of near real-time data processing, you only need to make sure that your app uses a Performance Monitoring SDK version that's compatible with real-time data processing.

These are the real-time compatible SDK versions:

• iOS — v7.3.0 or later
• tvOS — v8.9.0 or later
• Android — v19.0.10 or later (or Firebase Android BoM v26.1.0 or later)
• Web — v7.14.0 or later

Note that we always recommend using the latest version of SDK, but any version listed above will enable Performance Monitoring to process your data in near real time.

Which versions of the Performance Monitoring SDK are considered real-time compatible?

These are the SDK versions compatible with real-time data processing:

• iOS — v7.3.0 or later
• tvOS — v8.9.0 or later
• Android — v19.0.10 or later (or Firebase Android BoM v26.1.0 or later)
• Web — v7.14.0 or later

Note that we always recommend using the latest version of SDK, but any version listed above will enable Performance Monitoring to process your data in near real time.

What if I don't update my app to use a real-time compatible SDK version?

If your app doesn't use a real-time compatible SDK version, you will still see all your app's performance data in the Firebase console. However, the display of performance data will be delayed by roughly 36 hours from the time of its collection.

I've updated to a real-time compatible SDK version, but some of my users are still on old versions of my app. Do I continue to see their performance data in the Firebase console?

Yes! Regardless of which SDK version an app instance uses, you'll see performance data from all your users.

However, if you're looking at recent data (less than roughly 36 hours old), then the displayed data is from users of app instances using a real-time compatible SDK version. The non-recent data, though, includes performance data from all versions of your app.