Cortex, the open source, horizontally-scalable, highly-available, clustered Prometheus implementation that powers Grafana Cloud’s Hosted Prometheus cut its first release yesterday! This release was led by Chris Marchbanks from Splunk, who put together the whole release process and shepherded this first release:
We’re hoping to release every 4-6 weeks from now on. As Bryan Boreham, one of the Cortex maintainers, said, “There’s no particular new feature to celebrate, but lots of folks asked us for stamped versions so… you got it!”
I started the Cortex project over 3 years ago. Cortex has been in production for almost as long; we never followed a specific release cycle, preferring instead to continuously deploy the latest build of the master branch to our staging environment for testing, promoting the build to production when we were satisfied. While this system suited people working on Cortex day-to-day, it became clear that the vast majority of users would prefer versioned releases.
Cortex is now in production usage at many companies, including EA, Aspen Mesh, Mayadata, and Gojek. What’s more, the distributed systems data structures and algorithms that underpin Cortex help us accelerate the development of Loki, Grafana’s log aggregation system.
I could never have imagined the growth Cortex has seen. Cortex was accepted into the CNCF Sandbox almost a year ago, and has received contributions from more than 50 people! I want to take this opportunity to send a massive thank you to everyone who has contributed to Cortex and helped build the Cortex community.
What does the future hold for Cortex? At Grafana Labs, we continue to push the scale of our Cortex clusters, now handling many millions of samples per second. We’re working on using Prometheus’s TSDB in Cortex to make it easier to run on premise. We’re improving the cluster joining and leaving process so we can dynamically scale Cortex clusters on demand. And we’re bringing some of Cortex’s PromQL acceleration techniques to vanilla Prometheus and Thanos.
With the recent increased interest in building distributed timeseries databases, now is a very exciting time to be working on Cortex!