Alibaba Cloud ACK Cluster

alicloud-ack-cluster

A production-ready Alibaba Cloud ACK managed Kubernetes cluster (Pro tier) wired into an existing VPC, with autoscaling worker node pools. Designed to sit on top of a network foundation such as alicloud-vpc-foundation: you bring the private vSwitches, this module brings a hardened cluster.

Status: static-validated, live-test pending. Validated with tofu validate + tflint + checkov against the aliyun/alicloud provider. Not yet applied against a live Alibaba Cloud account (no sandbox subscription, and ACK clusters take ~15 min to create and bill a per-cluster fee), so it ships under live-test quarantine.

Design & secure defaults

• Private API server. slb_internet_enabled = false by default — the Kubernetes API server is reachable only from inside the VPC (use a bastion, VPN, or Cloud Enterprise Network). Opt into a public endpoint explicitly.
• Pod-scoped RAM identity (RRSA). enable_rrsa = true turns on RAM Roles for Service Accounts: pods assume narrowly-scoped RAM roles via OIDC instead of sharing the node's instance role — least privilege for workloads. (Pro only.)
• Deletion protection on. Accidental cluster deletion is blocked; a Terraform destroy requires flipping deletion_protection = false first.
• Secrets envelope encryption. Pass encryption_provider_key (a KMS key id) to encrypt Kubernetes Secrets at rest in etcd. (Pro only.)
• Advanced security group (is_enterprise_security_group = true), the recommended type with a higher rule ceiling.
• Encrypted worker disks. Node-pool system disks default to encrypted cloud_essd (PL1); supply system_disk_kms_key for a customer-managed key.
• Control-plane audit logging. apiserver/kcm/scheduler/ccm logs ship to Log Service (SLS) with a 30-day retention by default.
• Managed node pools. Auto-repair and auto-upgrade are on so unhealthy nodes self-heal and patch upgrades roll out within the maintenance window.
• No NAT by default. new_nat_gateway = false — reuse the egress your VPC foundation already provides instead of spawning a second NAT gateway/EIP.

Cross-field invariants (Terway needs pod vSwitches; RRSA and Secrets encryption need a Pro cluster; autoscaling needs min_size <= max_size) are enforced with precondition blocks, so misconfigurations fail at plan time, not mid-apply.

Provider

aliyun/alicloud >= 1.241.0, < 2.0 (floor raised so the non-deprecated vswitch_ids argument is available). Requires Terraform/OpenTofu >= 1.6.

License

Commercial — LicenseRef-IaCBazaar-Commercial. See the IaC Bazaar terms.