Prometheus v3.11.0 released on 02-04-2026
Prometheus v3.11.0 is out now. It brings a raft of service-discovery changes and deprecations (Hetzner, AWS, Azure, Vultr, Consul, Kubernetes), a new SD metric to track update timestamps, and several TSDB controls and experimental features that affect storage and startup behaviour.
See the Prometheus GitHub release notes or the project’s official pages for full details and upgrade instructions.
What’s in this release
- Hetzner service discovery: role-specific label renames and deprecations, with some hcloud labels set to stop working after 1 July 2026.
- Discovery metric: new prometheus_sd_last_update_timestamp_seconds to track when SD updates were last sent to consumers.
- TSDB and storage controls: storage.tsdb.retention.percentage to cap disk usage, removal of stale temporary checkpoints, and experimental flags (fast-startup writing series_state.json, st-storage, xor2-encoding).
Upgrade notes
- Hetzner deprecations: update configs to the new labels (for example __meta_hetzner_robot_datacenter and the renamed hcloud location labels) before 1 July 2026 to avoid lost labels.
- TSDB and feature flags: retention handling and I/O behaviour changed — Prometheus now falls back to CLI flag values when retention is removed from the config, and use-uncached-io will fail early on unsupported environments; check retention settings and experimental flags when upgrading or downgrading.
Share any experiences or issues you see while upgrading — practical notes from the field help other users decide what to test first.