Prometheus Public API Service Limits
This document outlines service limits for query endpoints available through Sysdig Monitor’s Prometheus-compatible API. These limits are in place to ensure consistent performance and availability across all tenants
Service Limits
The following limits apply to Prometheus query endpoints in Sysdig Monitor:
| Limit Category | Value | Applies To | Notes |
|---|---|---|---|
| Maximum number of different metrics a query can scan | 10 | /query, /query_range | A query that scans more than this number of different metrics (e.g. using a regex with the name label) will be rejected. |
| Maximum number of samples returned per time series | 1500 | /query | If the combination of the start, end and step parameters results in more than this number of samples per time series, the query will be rejected. |
| Number of time series returned for queries not using PromQL aggregations. | 1000 | /query, /query_range | If a query is not leveraging any PromQL function (e.g. max(), avg_over_time()), the number of returned time series will be limited to this number. A warning message will be returned as a part of the response. |
| Maximum resulting segments | 100k | /query, /query_range | For queries that utilize an aggregation operator (e.g. avg()), the number of segments returned will be capped at this number |
| Maximum concurrent queries | 200 | /query, /query_range, /series | Excess queries may be queued or rejected. |
| Maximum rate of samples scanned per minute | 4 Billion samples per minute | /query, /query_range | Queries after this rate has been exceed will be rejected. |
| Query timeout | 2 minutes | All query endpoints | A global timeout for all queries. |
Best Practices
To ensure stable performance and avoid hitting service limits:
- Scope queries narrowly using label filters to reduce the number of returned time series.
- Limit time ranges to avoid querying wide historical windows unless needed.
- Use appropriate step values to reduce result size when querying large time ranges.
