[Arm64 Embedded board, offline single node cluster]: Kubernetes Memory requirement

Asking for help? Comment out what you need so we can get more information to help you!

Cluster information:

Single Node cluster on offline arm64 Embedded target board
Kubernetes version: v1.20.2
Cloud being used: ARM64 Embedded Target
Installation method: offline target, manually using v1.20.0 arm64 binaries.
Host OS: Linux 4.14
CNI and version: calico v3.9.6-arm64
CRI and version:


Created single node cluster on Arm64 Embedded target board in offline.
Using cross build kube tools for arm64 and prebuilt docker images of kube from v1.20.0 release.

With all arm64 docker images, (proxy, controller-manager, apiserver,scheduler, etcd, coredns, pause)
And calico v.3.9.6-arm64 docker images, created a single node cluster and node is in Ready state.

Memory used by basic kube tools and docker images of kube-system is ~720MB.

Almost 2.5GB of memory used as cache for page cache, inode & dentries cache by linux/system when docker images loaded on target and kube tools are executed for cluster init.

Can you please help with minimum RAM memory required for single node offline cluster on Arm64.

Are there any ways to reduce the memory footprint (Already stripped the kube tools, dynamically linked).

Found etcd docker image
k8s.gcr.io/etcd:3.4.13-0 312MB has previous revisions image, which creating more overlay fs entries.