Have you heard of a new tool that automates right-sizing your resources for Kubernetes and dares to be “the best VPA not to waste memory”? Meet Kondense:

* It auto-scales pods based on memory pressure, meaning all cold/unused memory pages are continuously removed.
* Technically, it runs as a sidecar and resizes containers in its pod to facilitate the required memory pressure. Every second, all unused memory is taken away while preventing out-of-memory errors.
* This tool's memory resize algorithm is based on Meta's Transparent Memory Offloading (TMO).
* While it’s focused on memory, CPU resources are resized, too (based on CPU usage).
* It works for Kubernetes clusters that run on Linux only; containerd version should be 1.6.9+, and the container’s Linux kernel should be 4.20+.

➡️ GitHub repo
📣 Reddit announcement

#tools