Memory Store Observability

The memory stores observability dashboard provides real-time charts on your memory usage and API requests. It also has a built-in alert system that notifies you by email when an issue arises to help you troubleshoot in sync. For further information about specific errors, you can view your Error Report to find the error logs.

Accessing the Dashboard

The memory stores observability dashboard is available for any experience using MemoryStoreService, but you must either be the experience owner or have analytics group permissions to access the dashboard.

To access the dashboard:

  1. Navigate to the Creations page on the Creator Dashboard.
  2. Under the Creator Hub dropdown, select your account or the group owning the target experience.
  3. Select the experience.
  4. In the Monitoring dropdown, select Memory Stores.

Available Charts

The dashboard includes two categories of line graphs:

  • Quota Usage charts for tracking your usage compared to the dynamically allocated quotas, which are calculated based on the number of users in your experience.
    • Memory Usage on your memory usage per minute in bytes compared to how much your allocated quota left.
    • API Request Unit on your total request units per minute compared to how much your allocated quota left. This chart can be broken down by each API method.
  • API Usage and Performance charts for monitoring the API usage pattern and performance based on API method and response status.

The chart contains data for the past 30 days, and you can select to view a custom time range with the selector at the top of the page. If you select a time range earlier than 30 days, the system returns a Request Failed error.

Response Status Codes

The Observability Dashboard's Request Count by Status and Requests by API x Status charts include status codes of API responses that you can use to understand and troubleshoot errors. For a table that lists and describes all of these status codes, see Observability. In addition, for information on how to resolve these errors, or the specific quota or limit that an error relates to, see Troubleshooting and Limits and Quotas, respectively.

Notification Alerts

The email alert system automatically detects and sends you alerts of memory store usage issues that can harm the performance of your experience. There are two types of alerts:

  • Warnings are alerts with lower priority about issues that can potentially impact your experience. You should resolve them before they become critical issues. The dashboard flags these alerts by highlighting the chart in yellow with a warning message.
  • Criticals are alerts with higher priority on issues directly impacting your experience. You should resolve these issues as soon as possible to prevent degraded performance. The dashboard flags these alerts by highlighting the chart in red with an error message.

When your experience triggers two types of alerts at the same time, the dashboard displays the alert header message and highlights the chart region for your attention as Critical.

Currently, the system has four built-in alerts for memory usage and API requests. If your experience exceeds an alerting threshold, the system notifies you at most once a day for each alert. When receiving an email alert, you should view the dashboard and take actions to prevent degraded performance.

Memory Usage Alerts

The system sends the following memory usage alerts:

  • A Warning alert when your experience's memory usage has exceeded 70% of the total quota within the past hour.

  • A Critical alert when your experience exceeded the memory size quota for the past hour.

These issues usually occur when you overpopulate a memory store data structure with too much data, which can consume the quota quickly even if you only send requests on a few keys. For example, if you set a long expiration time on keys or don't clean up processed data, extra data can fill up your memory quota.

To prevent and resolve these issues, apply best practices to keep track of data structures you're actively using and set up a mechanism for cleaning up processed data.

API Request Alerts

The system sends the following API request alerts:

  • A Critical alert when the number of your memory store request failures has exceeded 20% within the past hour.

  • A Critical alert when more than 10% of your memory store requests are being throttled for the past hour.

    • This alert is based on the number of DataStructureRequestsOverLimit + TotalRequestsOverLimit, which are error responses indicating that your experience is throttled because it sends too many requests.
    • This alert isn't related to the request size, only quantity.

    All API requests have corresponding response status codes, so you can view the ratio of Success responses to all error responses using the Request Count by Status chart for troubleshooting.