Prometheus v3.16.0 released on 05-11-2026
Prometheus v3.16.0 is out now. It brings several PromQL improvements and experimental features, plus robustness fixes to TSDB, WAL handling and remote write/read pipelines that will matter for reliability and observability accuracy.
See the Prometheus GitHub release notes for full details and upgrade instructions.
What’s in this release
- PromQL: an experimental integral() range-vector function, experimental query cost estimation and reporting with configurable execution limits (enable with –enable-feature=query-cost), and smoothed and anchored vector selectors made stable.
- TSDB and WAL: exemplars with non-unique label names are now rejected; TSDB preserves mapped chunks and series identities for exemplar recovery after compaction/restart; multiple fixes prevent loss of out-of-order samples during WAL checkpoints and compaction races.
- Remote write/read and OTLP: Remote Write 2.0 stops exposing several deprecated metadata metrics and now counts samples/histograms ingested without metadata; OTLP receiver and OpenMetrics 2.0 fixes improve exemplar handling and histogram behaviour.
Upgrade notes
- Breaking change: exemplars with duplicate label names are rejected. Check any instrumentation or exporters that attach exemplars and ensure label names are unique before upgrading.
- Rollback: if you must revert, restore the previous Prometheus binary and data files and verify TSDB/exemplar compatibility before returning the instance to production.
Tell us how the upgrade went — share any unexpected behaviour or improvements you spot when running queries, remote writes or TSDB operations.
