Prometheus 2.16.0 is now out, following on from 2.15.0 with many fixes and improvements.

There's more performance improvements in this release. If you have queries that only touch the head (most recent hour-ish of data), they will be faster as I've removed some some unnecessary sorting. In practice this should make recording rules (which should generally only be working on recent data) about one second faster if they were working with a high cardinality of 1 million series.

It's now possible to configure Prometheus to log all PromQL queries it performs.

The PromQL parser has been replaced with a generated parser with improved error messages, which also happens to be a bit faster. There's also a new absent_over_time function. Reloads of rule groups and scrape targets are more efficient where nothing has changed.

There's new prometheus_target_metadata_cache_* metrics for the memory taken by target metadata, and prometheus_tsdb_wal_writes_failed_total for WAL write failures.


There’s also a number of improvements, bug fixes and other changes, see the release notes for more detail.


Wondering how to upgrade? Contact us.