Monitoring system and time series database Prometheus is now available in version 2.12.
The release comes with new features such as logging for additional components, enhancements and bug fixes supposed to improve usability.
Prometheus 2.12 is the first version to track active PromQL queries in a log file. Queries are written into a mmaped file in JSON format and removed once done. Should the software crash, the queries active at the time of the failure are printed on stdout in a rerun.
To help builds on mips64/mips64le targets, the Prometheus team has updated fsnotify to a more current version that should include all the system calls needed. It has also changed the resharding calculations slightly, to resolving some of the issues people had experienced tuning shards.
Smaller enhancements include a more precise flushing of TSBD pages and improved responsiveness of the web UI and API endpoint. The latter was realised by allowing the targets to be run in parallel to limit blocking time.
Other than that the team has fixed some bugs that lead to an incorrect handling of empty labels from alert templates and faulty error propagation in WAL watcher helper functions. An additional function is supposed to help with mergeSeriesSet problems by checking for and removing duplicate label names in remote reads. The complete release notes can be found in the project’s GitHub repository.
Prometheus was built at SoundCloud and joined the Cloud Native Computing Foundation in 2016 in an attempt to clarify the project’s independence. It graduated from the CNCF’s incubation program in August 2018, making it the second project to reach this status, following container orchestrator Kubernetes.