# Small footprint ~~TKG~~ TKR for VeloCloud

### Requirements
- No Cluster API
- no management cluster
- But there is a cluster representation
- TMC / management cluster will reconcile this cluster view
- Small footprint
- kapp is pretty big
- everything in TKG is really big
- Needs to integrate with TMC
- Single node cluster
- why k8s? its just kubelet
- upgrades? maintenance? backup / restore

### scope it for velocloud
- min k8s
- TMC

## Install
- Use same `host-agent`/installer
- If mode is non clustered, generate a kubeadm config
- Download & install all components locally
- Install TMC agent (would have to be smaller footprint one)

## Upgrades
- Same as BYOH. Uninstall and install.
- Should be ok, since workloads deployed via TMC

## HA?
- No HA, since single device
- No clustering of devices - yet.

## Performance constraints
- Does velocloud device have fast enough storage for etcd?
- How small do we need to be? Is shrinking kubelet + combining binaries enough?

## To reduce footprint
- Combine binaries to run
- Compile kubelet without cloud providers
- Anything else?
- Reduce TMC agent footprint (Erol has some work done)
- kubeadm would need to change. Wont happen as easy. compare existing k8s with

## What is TKG about it?
- Uses TKR bits
- Should hopefully be the same as a TKGm cluster, so no added support burden for TMC

# Out of scope
- Since single node cluster, no need to worry about low latency, poor connectivity use cases