Monitor invocation metrics

Invocation metrics are indicators of the outcome of a Forge app function invocation. When monitoring invocation metrics for your app, we recommend using filters to refine the results.

To view invocation metrics:

1. Access the developer console.
2. Select the Forge app that you want to view metrics for.
3. Select Metrics in the left menu.
4. Select Invocation in the left menu.

The image below shows invocation metrics, as well as all sites that your Forge app is currently installed on, and where there has been at least one invocation in the selected time period. If there hasn't been any invocation, the charts won't show any data.

Invocation metrics

The following metrics are available for all function invocations.

• Invocation success rate: The percentage of successful vs. failed invocations, across all functions. An invocation is considered successful if the function doesn’t fail with an invocation error.
• Invocation count: The total number of invocations, regardless of success or failure.
• Invocation errors: The number of invocations that failed with an error. They are grouped into the following error types:
  • Out of memory: The function has exhausted the available memory.
  • Timeout: The function has not been successful within a time limit.
  • Deprecated Runtime: The function is using a deprecated runtime. We recommend that all apps upgrade to the current Node.js runtime. If needed, you can backport minor version updates to older major versions of your app.
  • Unhandled exception: The function threw an uncaught exception. This category may include exceeding other platform limits such as network requests. To learn more about why the exception was thrown, view your app logs by selecting Logs in the left hand menu. For more information, see View app logs.
• Invocation response time: The time it takes for an invocation to successfully complete. The charts in this section show the distribution of invocation time as a histogram across different time buckets.

The Invocation response time section may have one or two charts, depending on whether your app uses Forge Containers (EAP):

• Remote and function invocation response time: displays the invocation time for each function under the handler field in the app manifest.yml file. This invocation time is measured from inside the AWS lambda, and doesn't include cold start, but it includes the time it took for the lambda initialization phase to complete.
• Container invocation response time (only if your app uses Forge Containers): displays the response time for each
  container defined in the app manifest.yml file.

Invocation errors

To learn more about your app's invocation errors, select the chart title, or select the More actions (⋯) menu on the chart and View details.

The following screen appears, showing site-specific information about your app's invocation errors.

In this view, you can search, filter, and sort the data to identify errors across specific sites and installations. You can also group the chart by version and error type, by selecting the Group chart by dropdown above the chart.

Invocation time

To learn more about your app's invocation time, select the chart title, or select the More actions (⋯) menu on the chart and View details.

The following screen appears, showing function-specific information about your app's invocation time.

In this view, you can search, filter, and group the data to see invocation time across specific functions, environments, and time periods.

You can also see a summary of the following percentiles involving the invocation time of the Forge app:

• P50 - Median
  • Indicates the value of the response time that's faster or equal to 50% of the app's speed of function invocations.
  • This is the typical performance of your app's speed of function invocations and is not skewed by extreme values.
• P90 - 90th percentile
  • Indicates the value of the response time that's faster or equal to 90% of the app's speed of function invocations.
  • If the P90 value is 170 ms, this means that the response times of 90% of the requests your app receives is less than or equal to 170 ms.
  • This helps give an understanding of what the slowest 10% of users may be experiencing with their response times.
• P95 - 95th percentile
  • Indicates the value of the response time that's faster or equal to 95% of the app's speed of function invocations.
  • If the P95 value is 170 ms, this means that the response times of 95% of the requests your app receives is less than or equal to 170 ms.
  • This helps give an understanding of what the slowest 5% of users may be experiencing with their response times.

To group the chart by function or version, select the Group chart by dropdown above the chart. Note, this will only reflect in the chart data, not in the table data.

Filters

Use these filters to refine your metrics:

• Environment: Narrows down the metrics for a specific app environment for your app.

• Date: Narrows down the metrics based on your chosen time interval. Choose from a range of predefined values, such as the Last 24 hours, or choose a more specific time interval using the Custom option.

• Sites: Narrows down the metrics based on the sites that your app is installed onto, for example, <your-site>.atlassian.net. You can select multiple sites.

• Atlassian app: Narrows down the metrics to show data from a specific Atlassian app if your Forge app supports multiple Atlassian apps.

Each metric can also be grouped by app version or function, by selecting the Group by dropdown next to a chart.

You can bookmark the URL on your browser to access metrics based on specific filtering criteria for quick access.