The team behind monitoring system and time series database Prometheus has pushed version 2.28 of its project out into the open. After some months of testing, and some last bug fixes to make sure autocompletion and parsing works as intended for all sorts of float values, the release is the first to sport the PromQL editor as the default.
The editor was added into the open-source project with the goal of providing advanced autocompletion, syntax highlighting, and linting capabilities for the project’s querying language. It is now deemed ready for general consumption.
While this is probably the most obvious change in the user interface, it isn’t the only one . Prometheus’s graph pages also received some love and are now able to render exemplars on top of time series — a feature inspired by the Grafana visualisation tool (and team).
When data is imported or backfilled, Prometheus’s promtool can now be prevented from displaying which data was added by setting a --quiet flag. Other than that, the update comes with an experimental body_size_limit for target scrapes, intended to protect the Prometheus server from running into memory issues.
Prometheus 2.28 also includes various improvements in the service discovery area, which for example adds Linode and generic HTTP-based service discovery, namespace support for Consul Enterprise in Consul SD, and an ingress class name label for Kubernetes SD. The updated project will also show a startup screen with a progress bar to give users a better idea of when the time series database (TSDB) will be ready.
TSDB itself has been fitted with better validation of exemplar label set length, and a new metric indicating whether a TSDB lockfile from a previous run still existed on startup. It also learned to split chunks, if they would contain more than the maximum 120 samples after being merged.