Adaptive Polling

This document describes the behavior, configuration, and practical use of the adaptive polling feature in the AWS CloudWatch integration for Splunk Observability Cloud.

Overview

Adaptive polling helps reduce AWS CloudWatch costs associated with the Splunk Observability Cloud integration by decreasing the polling frequency for recently unused metric data.

Adaptive polling in AWS CloudWatch integration

In adaptive polling mode, the integration continuously monitors which AWS CloudWatch metrics are actively used. Based on recent access patterns, metrics are classified as either active metrics or inactive metrics.

  • Active metrics are metrics that have been recently accessed, for example viewed in charts or retrieved through SignalFlow API calls.
  • Inactive metrics are metrics that have not been accessed recently.

The integration applies different polling intervals to these categories:

  • Short polling interval for active metrics
  • Long polling interval for inactive metrics

Using different polling intervals reduces CloudWatch API consumption while maintaining visibility into the metrics that matter most.

Technical criteria for determining active metrics

A metric is considered active up to 3 hours after it was last accessed.

Notes:

  • A metric displayed in a chart should be treated as unused unless the chart itself is actively viewed.
  • When someone opens a chart containing a metric, or accesses the metric via the API, AWS CloudWatch data is synchronized from the last time it was fetched. This typically occurs within up to 5 minutes.
  • When metric status changes to inactive Splunk Observability Cloud applies the longer poll rate.

Before you start

To use this feature, configure two distinct AWS CloudWatch polling intervals:

  • Active metrics: 1 to 10 minutes
  • Inactive metrics: 1 to 60 minutes (Inactive metrics polling interval must be longer than Active metrics polling interval)

Using the API

In addition to the standard pollRate field, set the inactiveMetricsPollRate field in the integration. This field defines the poll rate for unused metrics, in milliseconds.

For more details, refer to Splunk documentation on configuring the AWS CloudWatch integration via the API.

To enable Adaptive Polling at scale for existing AWS CloudWatch integrations, you can use the Python automation script available in the Splunk Adaptive Polling repository.
CODE 
python enable_aws_adaptive_polling.py \
  app.us0.signalfx.com \
  <API_TOKEN> \
  --inactiveMetricsPollRateMinutes 15

Using the Splunk Observability Cloud UI

  1. Create a new integration or edit an existing one.
  2. Navigate to the Define AWS Connection view.
  3. Enable CloudWatch Metrics, and then enable Adaptive Polling.
  4. Configure the two poll rates described earlier.
For more details, refer to the documentation on connecting Splunk Observability Cloud to AWS using polling.

Viewing charts with inactive metrics

If a chart displays outdated or stale metrics, it means those metrics are in a not‑recently‑used state. Try refreshing the chart after approximately 5 minutes — the time needed for metrics to warm up and synchronize.

Miscellaneous

If you begin using an AWS CloudWatch metric in a chart or via the SignalFlow API and new data does not appear promptly, consider setting a concrete value for the max delay parameter.