#Huawei DCS Provider Release Notes

#Overview

This page tracks release notes for the Alauda Container Platform Huawei DCS Infrastructure Provider. Each section corresponds to a release tag. Releases are ordered newest first.

For the current full capability list and installation steps, see Huawei DCS Provider Installation.

#v1.0.18 (2026-05-19)

#New Features

• Safer VM deletion lifecycle — The provider now stops a virtual machine before issuing the delete operation against Huawei DCS. The provider waits for the VM to reach a stopped state, then performs the delete; if the VM is already gone the operation completes idempotently. This prevents orphaned VM state in scenarios where the delete API rejects or retries against a running VM.

• retainVm annotation on DCSMachine — Setting the retainVm annotation on a DCSMachine short-circuits both the stop and the delete steps. The DCSMachine finalizer is released cleanly and the underlying VM remains running in DCS. Useful when a VM should be detached from the cluster for offline inspection or migration without being destroyed.

• VMStopPending status condition on DCSMachine — A new status condition reports whether a node deletion is currently waiting for its VM to stop. The condition is observable through kubectl describe dcsmachine and integrates with controller logs, providing visibility into the stop-then-delete lifecycle stage.

#Enhancements

• Etcd encryption restoration during cluster reconciliation — Etcd encryption configuration and datastore handling are restored consistently during reconciliation of existing clusters, improving recovery behavior after control plane churn.

• Etcd synchronization for the global cluster — The provider now correctly handles the global cluster during etcd synchronization, aligning behavior with workload-cluster handling.

• FC/DCS domain user support in cluster credentials — The DCSCluster Secret reference accepts a userType field. Domain users on Huawei FC and DCS are recognized through the new semantic Secret key. Local administrator accounts continue to work as before.

#Fixed Issues

• DCSMachineTemplate system disk size validation — The system disk size declared in DCSMachineTemplate must be at least equal to the source VM template's system disk size; the controller now reports the explicit error sooner during reconcile.

#v1.0.17 (2026-04-29)

#Enhancements

• Cluster Plugin UI image tag refresh — Internal management UI image tag updated. No user-visible behavior change in the provider controller.

#v1.0.16 (2026-04-29)

#New Features

• Pool-managed persistent disks on Huawei DCS — Persistent disks are now declared and tracked through the DCSIpHostnamePool resource. Declared persistent disks can be detached from old VMs and reattached to replacement VMs during rolling upgrades, reducing the risk of data loss on preserved node-local disks.

• DCSIpHostnamePool.spec.pool[].persistentDisk and persistentDiskStatus — DCSIpHostnamePool carries persistent disk specifications and runtime tracking through the new persistentDisk and persistentDiskStatus fields. Persistent disks are declared through YAML manifests only; this configuration is not exposed in the web UI.

• Platform-required /var/cpaas disk — The /var/cpaas disk is now documented as a persistent disk configured on the IP slot. It is no longer declared inside DCSMachineTemplate.

#Upgrade Notes

Existing clusters can migrate to the pool-managed persistent disk model by following Migrate Existing Huawei DCS Clusters to Pool-Managed Persistent Disks.

#v1.0.15 (2026-04-13)

#New Features

• Etcd encryption configuration support — Initial implementation of etcd encryption configuration for clusters managed by this provider. Note: a more complete automatic generation and key-rotation flow shipped later in v1.0.9 of the next major track; treat this entry as the first DCS-side support for the capability.

#Fixed Issues

• Kube-OVN reconciliation timing — Kube-OVN reconciliation is now permitted when control plane nodes are still partially initializing, avoiding a chicken-and-egg condition during initial cluster bring-up.

#Enhancements

• imagePullSecrets on apiservice ServiceAccount — The apiservice ServiceAccount now accepts imagePullSecrets, enabling image pull from registries that require authentication.

• Internal toolchain refresh — Internal Kubernetes API client libraries updated to align with newer platform baselines.

#v1.0.14 (2026-03-06)

#Enhancements

• OpenTelemetry SDK upgrade — Bumped go.opentelemetry.io/otel/sdk to v1.40.0 for security and maintenance updates.

#v1.0.13 (2026-02-28)

#New Features

• DCS proxy API — Added a proxy API that exposes selected DCS platform operations through the provider, simplifying integration for callers that should not access the DCS platform API directly.

#v1.0.12 (2026-02-10)

#Enhancements

• Deployment storage limit configuration — Provider Deployment now declares ephemeral storage limits, preventing the controller pod from consuming unbounded node disk space.

#v1.0.11 (2026-02-09)

#New Features

• Dedicated CD-ROM datastore for ignition ISO upload — The DCS provider can now use a dedicated datastore for uploading the ignition ISO via the CD-ROM device, separating ISO storage from general VM datastores when policy requires it.

#v1.0.10 (2026-01-22)

#Fixed Issues

• CVE patches — Resolved multiple disclosed CVE issues in vendored dependencies.

#v1.0.9 (2025-12-31)

#New Features

• Multiple network interface card (NIC) support on virtual machines — Cluster nodes can be configured with multiple NICs, enabling separated control, workload, and storage networks on the same VM.

• Custom node metadata through MachineDeployment and KubeadmControlPlane — Node labels, annotations, and taints can be declared on MachineDeployment and KubeadmControlPlane resources. Available with platform version 4.2.0 and above.

#Enhancements

• Automatic etcd encryption configuration — Encryption configurations are generated automatically during cluster creation. After etcd key rotation, no manual intervention is required during control plane upgrades.

#Fixed Issues

• Multiple security issues addressed in this release.

#v1.0.8 (2025-12-29)

#New Features

• Automatic etcd encryption configuration (initial) — The provider auto-generates the apiserver encryption configuration during cluster bring-up. This is the first cut; the fully automatic key rotation and post-rotation control-plane upgrade behavior is described in v1.0.9.

#Fixed Issues

• CoreDNS configuration fix — Resolved a CoreDNS configuration issue in clusters created through this provider.

#v1.0.7 (2025-07-09)

#Enhancements

• Internal naming cleanup — Adjusted provider component naming for consistency. No user-visible API change.

#v1.0.6 (2025-07-04)

#Enhancements

• DCSIpHostnamePool deletion annotation — The provider now sets a deletion annotation on DCSIpHostnamePool resources during teardown, improving the deletion-flow auditability.

#v1.0.5 (2025-07-02)

#New Features

• Initial retainVm functionality for DCSMachine deletion — First implementation of the retainVm capability: deleting a DCSMachine can be configured to leave the underlying VM running in DCS instead of destroying it. This capability was later promoted to a formal annotation on DCSMachine in v1.0.18.

#v1.0.4 (2025-06-19)

#Enhancements

• DCSIpHostnamePool cleanup on DCSMachineTemplate deletion — When a DCSMachineTemplate is removed, the associated DCSIpHostnamePool is now released automatically.

• Removed bundled kubeadm component from the DCS plugin — The DCS provider no longer ships its own bundled kubeadm component; kubeadm bootstrap is handled through the platform's standard Cluster API kubeadm provider.

#v1.0.3 (2025-06-11)

#New Features

• Data disk support on DCSMachineTemplate — DCSMachineTemplate accepts additional data disks beyond the system disk, allowing per-template disk inventory declaration.

• dcsctl CLI utility — Initial release of dcsctl, a command-line utility for DCS-side operations (for example, manual VM stop and folder operations referenced by later releases).

#v1.0.2 (2025-06-05)

#New Features

• /etc/resolv.conf in ignition configuration — Ignition userdata now writes /etc/resolv.conf so cluster nodes pick up DNS settings at bootstrap without requiring post-boot remediation.

• Generated kubelet TLS material — The provider can now generate kubelet.crt and kubelet.key during node provisioning, removing a manual step previously required for the kubelet TLS bootstrap.

#v1.0.1 (2025-05-27)

#Enhancements

• Stabilization patch following the initial v1.0.0 release. No user-visible API changes.

#v1.0.0 (2025-05-09)

#Overview

v1.0.0 is the initial release of the Alauda Container Platform Huawei DCS Infrastructure Provider. The provider implements the Cluster API model for the Huawei DCS (also referred to as FC / FusionCompute) platform, enabling declarative Kubernetes cluster creation and management on top of DCS-managed virtualized infrastructure.

#Initial Capability Set

• Cluster lifecycle on Huawei DCS — Create, scale, and delete Kubernetes clusters that use DCS as the infrastructure backend.

• DCSCluster, DCSMachine, DCSMachineTemplate, DCSIpHostnamePool CRDs — Declarative API surface for DCS-based clusters, including reserved hostname and IP pool management through DCSIpHostnamePool.

• Ignition-based node bootstrap — Node bootstrap uses ignition userdata. By design, the DCS provider uses ignition exclusively as the bootstrap mechanism for DCS-managed VMs.

• Kubeadm-based control plane and worker bootstrap — Cluster bootstrap aligns with the upstream Cluster API kubeadm bootstrap/control-plane model.