Prometheus 2.15 Released, memory usage is reduced

December 23, the development team of Prometheus released the latest stable release Prometheus 2.15.0.

Prometheus is an open-source chronological database designed for monitoring.  Implemented with multi-dimensional data model, it enables regular monitoring of the system and service status, evaluating rules, displaying results, and handling alert triggers.  With single server node, it collects chronological data via pull model over HTTP.  The project operates under The Cloud Native Computing Foundation (CNCF).

Prometheus 2.15 is the latest point release of the version 2 series, first released in 2017.  By setting a limit to how much a storage (TSDB) block can store symbols, the memory usage of a block has been optimized.  As to compaction, memory footprint has been reduced with optimized buffer.

A new endpoint was added to expose per metric metadata.  Via V1 API, the endpoint offers metadata during the lifecycle of the attached target.

On the other hand, prometheus_sd_kubernetes_cache_* has been removed.  The reason of the removal was that as there are metrics (prometheus_sd_kubernetes_workqueue_latency_seconds and prometheus_sd_kubernetes_workqueue_work_duration_seconds), which record workqueue duration and latency as seconds, it wasn’t being used.   

There are numerous other small feature improvements made on query language and others.