Adaptive rules
Adaptive rules are in early access and are rolling out to selected accounts. To request access, contact your Coralogix account representative or Support.
Adaptive rules keep your metrics optimized based on how you actually use them. Instead of setting a tier on every metric by hand, you define rules that match metrics by usage and automatically route each one to the right tier. Series nobody queries can move to a cheaper tier where they stay available and return in full the moment a query reads them, so you save while they sit idle without losing fidelity. Metrics you are confident you do not need can be blocked instead, which stops their ingestion and cost going forward.
Coralogix bills metrics on your active time series: the genuinely unique series read in a rolling 60-minute window, not on resolution or redundant labels. Adaptive rules lower cost by right-sizing where low-value series live, not by dropping labels or reducing how often you scrape. You keep full-fidelity metrics and pay only for your true cardinality.
What you need
- Adaptive rules enabled for your organization.
- Permission to manage metric data.
How adaptive rules work
Adaptive rules are evaluated across every metric in your account, from top to bottom. The first matching rule sets the tier for that metric, so order matters: put the rules you want to win at the top.
Each rule matches metrics by a usage condition and assigns one of these tiers:
| Tier | What it does | When to use it |
|---|---|---|
| Medium | Keeps the metric in the standard tier at full fidelity and full cost. | Metrics you actively rely on. |
| Low | Moves the metric to a cheaper tier. It stays fully queryable and returns in full the moment a query reads it, so you save while it sits idle. | Metrics that are rarely queried but you still want available. |
| Block | Stops ingesting the metric going forward, so its volume and cost stop. History already stored is kept, and you can unblock it at any time. | Metrics you are confident you do not need. |
Lowering a metric's tier keeps its data: an idle series simply costs less until a query reads it again, with no loss of detail. Blocking is different: it stops ingestion going forward, so reserve it for metrics you are sure you do not need.
Changes to adaptive rules can take up to 24 hours to take effect across your metrics.
Rule conditions
Adaptive rules classify metrics using each metric's own usage record: its cardinality, dimensions, label variations, and which queries read it. Rules match on these usage signals:
| Condition | What it selects | Why it is a safe candidate to down-tier |
|---|---|---|
| Unused | Metrics not queried in the last 14 days. | Nothing has read them recently. |
| Bogus | Metrics whose queries run but never return any series. | They are referenced by name, but hold no data anyone uses. |
| Only broken queries | Metrics read only by queries that always fail. | No successful query depends on them. |
| Ad-hoc only | Metrics read only in ad-hoc queries, never by a dashboard, alert, recording rule, SLO, or API. | Nothing persistent depends on them, so lowering their tier will not break a saved asset. |
Coralogix suggests a Recommended tier for each condition, which you can accept or change. Start with adaptive rules to act across every metric at once, then fine-tune individual metrics in the table.
Fine-tune individual metrics
Adaptive rules set the baseline across the account, and you can still act on any single metric. In the Metric usage analysis table, each metric carries its own usage record, so you can block one in a click and unblock it in another, regardless of the rules.
Where adaptive rules fit among your cost levers
Adaptive rules are the automatic, usage-driven layer. Combine them with the other levers depending on the problem:
- Adaptive rules: automatically tier or block metrics by how they are used. Reach for this first to optimize the whole account at once.
- Metrics cost optimizer: manually block or unblock specific expensive metrics.
- Recording rules: aggregate high-churn dimensions before they are indexed, when a metric is useful but too granular.
- Scrape interval: resolution only; it changes stored samples, not billed series, so use it to manage volume, not cost.
Verify the impact
After the rules take effect, confirm the impact in Metric usage analysis: the active series and billed units of low-tier and blocked metrics drop, while the metrics you kept stay fully queryable. For the limits that govern ingestion and querying, see Ingestion and query limits.