Adaptive Metrics after hitting limit

I didn’t discover that Adaptive Metrics existed until after I went over my metric limit and started looking around for what could help me. Currently, though, the Adaptive Metrics area doesn’t show any recommendations and the update date on the recommendation generation is from today. If the problem is a day with no/little data, how can I get AM to make some recommendations based on a longer date range? If that’s not the problem, what could be wrong?

Those are only recommendations. Nothing stopping you to write own rules. E. g. check metrics with high cardinality and you can write&aply AM rule to drop/reduce cardinality for those metrics, which you don’t use/need.

1 Like

Ah, my question is whether I should expect to get recommendations after my ingestion has been throttled or stopped. What time range does AM look at for its analysis? I did confirm that metadata is being sent for my metrics, per the docs.

I’ll definitely experiment with writing and applying custom AM rules once I see what recommended ones look like, the effects of applying them, etc. Since I didn’t discover AM exists until now, I have no idea at all of what my options are. I’ve tried dropping some metrics in my OTEL collector, but I wouldn’t hate having an easier UI and API for that if that’s basically what AM can provide. :slight_smile:

Hi @MAW - the recommendation service runs every 24 hours and evaluates all metric names + labels that have been received. If there are metric samples being dropped due to account limits, those metric names and labels will not be available for the Adaptive Metrics recommendation service to evaluate.

The Cardinality Management dashboards are generated from the last 2 hours of metrics that have been received, so that could be a good way to inspect metrics with many label name pairs that are not being used. Dropping those with client side filtering will reduce total metric series and hopefully improve the Adaptive Metric tool’s capability of evaluating metrics ingestion. Note: the Cardinality Management dashboards will also only show successfully received metrics. For details about monitoring ingest issues, refer to Monitor Prometheus for ingest errors | Grafana Cloud documentation

1 Like