Grafana Cloud

Review and apply recommendations

Manage your rules using the UI, API, or Terraform. This topic focuses on how to manage your rules using the Grafana UI.

Adaptive Metrics splits this work across two pages:

  • On the Metrics page, you review your metrics and the recommendations made for them, and apply those recommendations. You don’t create or edit rules here.
  • On the Configuration page, you manage what you have configured yourself: exemptions, rules, which usage sources feed the recommendation engine, and an audit log of rule changes. Rules are created, edited, and removed on its Rules tab. Refer to Create rules.

On the Overview page, you can see the current status and when rules were last updated.

To manage rules using the API or Terraform, refer to Adaptive Metrics HTTP API or Adaptive Metrics Terraform provider.

Review recommendations using the UI

  1. Navigate to Adaptive Metrics.

  2. Navigate to the Metrics page.

    A list of your metrics and their recommendations appears.

Filter metrics

Use the following controls to narrow down the list:

  • Segment: Choose the segment whose metrics you want to review. The segment selector applies to most of the plugin, not only this page.
  • Search: Search metrics by name.
  • Awaiting review: Toggle to show only the metrics whose recommendation you haven’t acted on yet.
  • Urgent: Toggle to show only the recommendations marked urgent.

If you have permission to write rules, a row below the filters holds the selection controls: a Show selected toggle, which is unavailable until you select at least one metric, an N selected count, and a button to clear the selection once you’ve selected something.

Note

There’s no filter by recommended action on the Metrics page. Use Awaiting review and Urgent instead. Match type filters the rules list on Configuration > Rules.

Metrics table

For each metric, the table shows the following columns:

  • Metric: The metric name.
  • Received series: The number of active series received for this metric.
  • Current reduction: The active series that the currently applied rule removes. The cell shows a dash when the figure isn’t known.
  • Actions: An Urgent badge when the recommendation is urgent, the projected change the recommendation would make, a button to apply the recommendation, and a Details button that opens the metric details drawer.

A leading checkbox column for selecting metrics appears if you have permission to write rules.

Select a column header to sort by it. Most columns sort one way only: Metric sorts A to Z, and Received series and Current reduction sort highest first. Those headers become unavailable once they’re the active sort, because there’s no second direction to switch to. The Actions header sorts by the projected change and is the only one you can reverse, so you can put either the biggest projected reduction or the biggest projected increase first. The list starts sorted by Received series, highest first.

Note

While a search is active, the list is returned in metric name order only, so the sort you chose doesn’t apply until you clear the search.

View metric details

Select anywhere on a metric’s row, or the Details button in the Actions column, to open the metric details drawer. The drawer shows:

  • Series Impact: Three cards showing Received (series ingested for this metric), Stored (series stored after the recommended rule is applied), and a combined Reduction card showing the absolute series change and percentage. The card shows Increase instead of Reduction when the recommended rule would add series. If no series are detected for this metric in the current segment, the drawer shows a message in place of the cards.

  • Metric usage stats: Three cards that always display the count of each source using this metric: Alerting and recording rules, Dashboards, and Queries. Zero counts are shown explicitly. Each card has a tooltip that explains how the recommendation affects that source. A recommendation can apply even when usage is nonzero, because it only changes parts of the metric that aren’t observed in use, so existing alerts, dashboards, and queries keep working.

  • Rule: A summary section describing the recommended action.

    • A banner at the top describes the recommendation in plain language. For Keep, it confirms no changes are needed. For Add, Update, or Remove, it describes the change and, when available, how stored active series would change.
    • A series impact badge next to the action buttons shows Decrease ~N series, Increase ~N series, No impact on series count, or Impact unknown, depending on the recommendation.
    • An Aggregation card shows the Types, Aggregate labels, Keep labels, and Managed by fields, when applicable. For drop rules, the card indicates that the entire metric is dropped.
  • Exemptions tab: Shows how exemptions affect this rule. Depending on the metric, the tab can include an Applied label exemptions table, a Recommendations muted list, a Rule suppressed by exemption warning, or a Rule effectively neutralized notice. If no exemptions affect the metric, the tab says so.

  • Dashboards tab: Lists the dashboards that use this metric. Each row shows the dashboard title and a button to open the dashboard in a new tab. The list paginates, and a search box appears when there are many dashboards. When dashboard links aren’t available, for example for system or internal dashboards, the tab notes this. If the metric isn’t used in any dashboards, the tab says so explicitly.

From the drawer, you can apply the recommended action by clicking Apply recommendation. A confirmation dialog box opens. The button doesn’t appear for Keep recommendations, or when you don’t have permission to apply recommendations.

Note

This drawer doesn’t edit rules or copy them as JSON. Both are on the drawer you open from Configuration > Rules, which carries a Rule badge so the two aren’t confused.

Apply recommendations

Before you apply, the Metrics page summarizes the segment above the apply controls: how many recommendations are awaiting review, the net series change if you applied them all, and when the segment was last analyzed. The Overview summary shows the same estimated series decrease or increase.

You can apply one recommendation at a time, or several at once. Either way, a confirmation dialog box opens first.

To apply a single recommendation, select the apply button in that metric’s Actions column. The confirmation names the metric and what applying does to its rule.

To apply several recommendations:

  1. Select the metrics you want to apply, then select Apply selected. To apply every recommendation in the segment instead, select Apply all without selecting any metrics.

    A confirmation dialog box opens, stating how many recommendations it applies. For Apply all, it also notes that this covers every recommendation in the segment, including any the filters are currently hiding.

  2. To confirm your rule updates, click Apply.

Apply all is unavailable when you don’t have permission to apply recommendations, or when nothing is awaiting review, and its tooltip says why. Because auto-apply leaves nothing awaiting review, the button is normally unavailable while auto-apply is on for the segment. You can still add and edit custom rules. For more information, refer to Enable auto-apply.

When Apply all changes at least one rule, Adaptive Metrics then offers to turn on auto-apply for that segment. The offer doesn’t appear after Apply selected, or when the apply changed nothing, and it returns the next time you apply all recommendations, until auto-apply is on.

After you apply rules, open Overview and check Time series reduction. Set the time range to include the apply and the period after it. Refer to Monitor and validate results.

Manage applied rules

Applied rules are listed on Configuration > Rules, not on the Metrics page. From that tab you can create a rule, and open a rule to edit or remove it. Refer to Create rules.

Removing several rules at once isn’t available on that tab.

Export rules

The Metrics page has no export control. To export your rules or recommendations, use the Adaptive Metrics HTTP API or the Terraform provider.