Unlock Elastic Scaling: Connect Your IDC Kubernetes to Alibaba Cloud ACK One Registration Cluster

Background and Challenge

Traditional IDC data‑center Kubernetes clusters are limited by physical capacity, making dynamic scaling difficult. ACK One registration clusters provide a cloud‑native way to extend resources on‑demand.

Advantages

Resource elasticity: Integrate on‑premises workloads with Alibaba Cloud resources for on‑demand compute expansion.

Cost optimization: Pay‑as‑you‑go pricing avoids sunk IDC costs.

Architecture Overview

The architecture connects an on‑premises Kubernetes cluster to an ACK One registration cluster, enabling unified management and cloud‑side elasticity.

Integration Steps

Create a registration cluster via console or CLI (see Alibaba Cloud documentation).

Deploy ack-cluster-agent on the IDC Kubernetes cluster.

Verify the agent is running:

kubectl -n kube-system get pod | grep ack-cluster-agent

Expected output shows the pod in Running state.

Creating a Cloud Node Pool

The node pool supports Elastic Container Instance (ECI), Elastic Compute Service (ECS), and GPU instances.

Prepare a custom script for GPU detection and driver installation:

#!/bin/bash
if ! which lspci &>/dev/null; then
  yum -y install pciutils
fi
if lspci | grep -i nvidia &>/dev/null; then
  echo "This node is a GPU node, install drivers."
fi

Configure the script in the registration cluster’s node‑pool settings.

Create the node pool and trigger scaling.

Verify new cloud nodes: kubectl get no Nodes named <RegionID>.<IP> are cloud‑provisioned ECS instances.

Deploying Pods to the Cloud Node Pool

Example deployment using a node selector:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment-cloud
  namespace: default
spec:
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      nodeSelector:
        alibabacloud.com/nodepool-id: np6d13af2bb56546128***
      containers:
      - name: nginx
        image: registry.cn-hangzhou.aliyuncs.com/eci_open/nginx:1.14.2
        ports:
        - containerPort: 80
        resources:
          limits:
            cpu: 500m

Scenario II – ACK Virtual Node (ECI)

Virtual nodes run Pods directly on Elastic Container Instances without managing ECS nodes.

Install the ack-virtual-node component via console or CLI.

Check deployment status: kubectl get no Virtual nodes appear with names like virtual-kubelet-....

Schedule a Pod to ECI by adding label alibabacloud.com/eci=true:

kubectl run nginx --image=registry.cn-hangzhou.aliyuncs.com/eci_open/nginx:1.14.2 -l alibabacloud.com/eci=true

Scenario III – ack‑co‑scheduler for Multi‑Level Elastic Scheduling

The ack-co-scheduler enables hierarchical scheduling: IDC resources first, then cloud node pool, then ECI.

Install ack-co-scheduler via console or CLI.

Create a ResourcePolicy to prefer IDC, then ECS node pool:

apiVersion: scheduling.alibabacloud.com/v1alpha1
kind: ResourcePolicy
metadata:
  name: scheduled-to-node-pool
spec:
  selector:
    app: nginx
  strategy: prefer
  units:
  - resource: idc
  - resource: ecs
    nodeSelector:
      alibabacloud.com/nodepool-id: np6d13af2bb...

Create a second ResourcePolicy to fall back to ECI:

apiVersion: scheduling.alibabacloud.com/v1alpha1
kind: ResourcePolicy
metadata:
  name: scheduled-to-eci
spec:
  selector:
    app: nginx
  strategy: prefer
  units:
  - resource: idc
  - resource: eci

Conclusion

ACK One registration clusters, cloud node pools, virtual nodes, and the ack‑co‑scheduler enable hybrid‑cloud Kubernetes deployments with elastic scaling, cost‑effective resource utilization, and automated multi‑level scheduling.