acs-virtual-node - Container Compute Service - Alibaba Cloud Documentation Center

ACS Virtual Node seamlessly connects Kubernetes with the computing power of Container Compute Service (ACS). This topic describes the components, usage instructions, and change log for ACS Virtual Node.

Component overview

ACS Virtual Node is based on the open source project Virtual Kubelet. It extends support for providers and includes many optimizations to seamlessly connect Kubernetes with the computing power of ACS. This gives Kubernetes clusters significant elasticity without being limited by the compute capacity of their nodes. You can create ACS pods on demand without planning your cluster capacity.

ACS Virtual Node runs application pods as serverless containers. This provides elasticity, eliminates the need for capacity planning, and offers on-demand usage and billing. In scenarios such as Jobs, CI/CD, Spark big data computing, and elastic online applications, it significantly improves the elasticity of application deployments and reduces computing costs.

Usage instructions

For more information about the ACS Pod feature, see Overview of ACS Pod instances.

Change log

January 2026

Version number	Change time	Changes	Impact
v2.16.2	January 28, 2026	Optimization: Optimized the retry mechanism for pod creation requests on virtual nodes when the quota is exceeded. Bug fix: Fixed an issue where pods on virtual nodes would occasionally get stuck in the Pending state because of unschedulable reserved nodes.	This upgrade does not affect your services.
v2.16.1	January 27, 2026	Optimizations: Optimized the internal HTTP connection pool management mechanism of the component. Optimized the pod creation performance of version v2.16.0 in large-scale scenarios to reduce startup time. Bug fix: Fixed an issue where pods would occasionally remain in the Pending state in rare cases.	This upgrade does not affect your services.
v2.16.0	January 06, 2026	Add The Pod Capacity feature is now enabled by default for virtual nodes. Added support for using the PodNetworking feature with ACS computing power. ACS computing power now supports referencing cross-namespace resources using `NodePublishSecretRef`. Added support for enabling the CSINode feature on reserved nodes. Reserved nodes can now send event notifications during self-healing. The `InterruptionHardwareFault` condition can now be displayed on pods that run on ACS GPU-accelerated instances and reserved nodes. Optimizations Optimized the component startup speed in large-scale scenarios. Improved the reliability of instance cleanup in edge cases. Bug fixes Fixed an issue where reserved nodes could enter the `NotReady` state in specific scenarios. Fixed an issue where pod conditions were unexpectedly overwritten in rare cases. Fixed an authorization issue that could occur when using SidecarSet injection with ACS computing power. Fixed an issue where the `Instance.ToBeReleased` condition was not correctly displayed on instances that use ACS computing power.	This upgrade does not affect your services.

December 2025

Version number	Change time	Changes	Impact
v2.15.8	December 08, 2025	Optimizations: Optimized the expiration of the local Informer cache. Improved the processing speed of pods in their desired state. Bug fixes: Fixed an issue where the permissions of SidecarSet pods were revoked. Fixed an issue where conditions were overwritten in rare cases.	This upgrade does not affect your services.

November 2025

Version number	Change time	Changes	Impact
v2.15.7	November 26, 2025	Optimizations: Optimized the rate-limiting retry mechanism. Bug fixes: Fixed an issue where the IP addresses of reserved nodes were not updated.	This upgrade does not affect your services.
v2.15.6	November 05, 2025	Optimizations: Optimized the resource normalization logic for ACS instances. Improved the error message for when a pod on a reserved node uses an unsupported `ComputeClass`. Bug fixes: Fixed an issue where the local Informer cache might not expire promptly. Fixed an issue where a pod's annotation could become inconsistent with its `Spec` definition after an in-place resource change (InPlace VPA).	This upgrade does not affect your services.

October 2025

Version number	Change time	Change description	Impact
v2.15.5	October 20, 2025	Bug fixes: Fixed an issue where the IP addresses of reserved nodes were missing in ACS clusters.	This upgrade does not affect your services.

September 2025

Version number	Change time	Change description	Impact
v2.15.4	September 28, 2025	Optimizations: Optimized the error message for DaemonSet pods scheduled to virtual nodes. Improved the recovery efficiency of GPU pods from abnormal states in ACS clusters. Bug fixes: Fixed an issue where the status of ACS pods was occasionally not updated. Fixed an issue where the IP addresses of reserved nodes in ACS clusters were missing. Fixed an issue where using CNFS persistent volumes occasionally caused crashes.	This upgrade does not affect your services.
v2.15.3	September 28, 2025	Bug fixes: Fixed an issue where pods remained in the `Pending` state for a long time in rare cases. Fixed an issue where pods could not reach their desired state (such as `Succeeded`/`Failed`) in rare cases.	This upgrade does not affect your services.
v2.15.2	September 22, 2025	Bug fixes: Fixed an issue in version v2.15.0 where ACS resource normalization did not take effect. Fixed an issue where the KubeletVersion field for reserved nodes was inaccurate. Fixed an issue where local cached data expired due to API Server push delays.	This upgrade does not affect your services.
v2.15.1	September 09, 2025	Bug fix: Fixed an issue where pod creation failed when a vSwitch was specified using an annotation.	This upgrade does not affect your services.

August 2025

Version number	Change time	Changes	Impact
v2.15.0	August 19, 2025	New features: Added support for the reserved node feature. Added support for automatic protocol conversion from ECI to ACS. Optimizations: Reduced program memory usage by about 40%. Improved the creation success rate of ACS pods. Adjusted the ephemeral storage capacity (`ephemeral-storage`) of virtual nodes to 60,000 TiB to prevent scheduling failures. Bug fixes: Fixed an issue where scaling out ACS GPU pods failed after a cloud disk was attached. Fixed an issue where in-place container restarts failed. Fixed an issue where the `PodNetworking` configuration did not take effect in ACK clusters.	This upgrade does not affect your services.
v2.12.0-acs.16	August 7, 2025	Fixed an issue where pods would occasionally remain in the Pending state in abnormal situations. Added support for reserved node status reporting. Optimized the reporting of the Interruption.GPUShareReclaim condition.	This upgrade does not affect your services.

May 2025

Version number	Change time	Changes	Impact
v2.12.0-acs.14	May 12, 2025	Corrected the unexpected update behavior of reserved nodes. Fixed a parsing error caused by an invalid format in the acs-profile configuration. Added support for exposing the Interruption.GPUShareReclaim condition.	This upgrade does not affect your services.

March 2025

Version number	Change time	Changes	Impact
v2.12.0-acs.13	March 20, 2025	Added support for inventory-aware scheduling for GPU pods.	This upgrade does not affect your services.

February 2025

Version number	Change time	Changes	Impact
v2.12.0-acs.12	February 13, 2025	Adjusted the ephemeral storage capacity on virtual nodes to 60,000 TiB. Fixed an issue where retries were not performed when a ServiceUnavailable error was encountered in advanced scheduling scenarios. Optimized the interaction with the csi-provisioner component when using persistent volume claims (PVCs) in late binding mode.	This upgrade does not affect your services.

January 2025

Version number	Change time	Changes	Impact
v2.12.0-acs.11	January 21, 2025	Optimized advanced scheduling capabilities.	This upgrade does not affect your services.
v2.12.0-acs.10	January 17, 2025	Added support for exposing metrics of reserved nodes. Optimized the retry mechanism. Other feature optimizations.	This upgrade does not affect your services.

November 2024

Version number	Change time	Changes	Impact
v2.12.0-acs.9	November 27, 2024	Added support for scaling out cloud disks for GPU pods. Added inventory awareness for General-purpose and compute-optimized pods. Other feature optimizations.	This upgrade does not affect your services.

September 2024

Version number	Change time	Changes	Impact
v2.12.0-acs.8	September 13, 2024	Upgraded the dependent Kubernetes API version to 1.30. Added the Prometheus metrics container_sockets and container_processes. Added support for network metric data collection using KubeSkoop. Added support for attaching cloud disks to GPU pods. Added support for hot-swapping CPUs for General-purpose pods.	This upgrade does not affect your services.

August 2024

Version number	Change time	Changes	Impact
v2.12.0-acs.7	August 21, 2024	Fixed an issue where associated events were missing a UID.	This upgrade does not affect your services.
v2.12.0-acs.6	August 02, 2024	Added support for scaling out cloud disk persistent volumes for General-purpose and compute-optimized pods.	This upgrade does not affect your services.

July 2024

Version number	Change time	Changes	Impact
v2.12.0-acs.5	July 25, 2024	Replaced Role+RoleBinding authentication with pod identity authentication.	This upgrade does not affect your services.

June 2024

Version number	Change time	Change description	Impact
v2.12.0-acs.4	June 21, 2024	Collaborated with the scheduler to support native Kubernetes scheduling semantics. Added support for creating pods using capacity reservation resources.	This upgrade does not affect your services.