acs-virtual-node

更新时间:
复制 MD 格式

ACS Virtual Node seamlessly extends Kubernetes with the container computing power of Alibaba Cloud Container Compute Service (ACS). This topic describes its components, usage notes, and release notes.

Component overview

Built on the open-source Virtual Kubelet project, ACS Virtual Node extends support for providers and has been significantly optimized to seamlessly integrate Kubernetes with the computing power of Container Service (ACS). This provides Kubernetes clusters with elasticity and removes node capacity constraints. You can create pods on-demand without planning cluster capacity.

ACS Virtual Node runs application pods as serverless containers, providing elasticity, eliminating capacity planning, and offering an on-demand, pay-as-you-go model. This model is ideal for significantly improving deployment elasticity and reducing computing costs for workloads such as batch jobs, CI/CD pipelines, Spark-based data processing, and handling traffic bursts for online applications.

Usage

For more information about ACS Pod, see ACS Pod Instance Overview.

Release notes

June 2026

Version

Date

Description

Impact

v2.17.3

June 5, 2026

  • Optimizations:

    • Reduced pod creation latency at scale.

  • Bug Fixes:

    • Fixed a rare issue where an instance could be deleted while waking up.

    • Fixed an issue from v2.17.0 where pods with long names were stuck in the Terminating state.

This upgrade has no impact on your workloads.

May 2026

Version

Date

Changes

Impact

v2.17.2

May 8, 2026

  • New Features:

    • Pod creation now automatically retries on an ImageCache.NotReady error.

  • Optimizations:

    • Increased the maximum number of metrics shards to 48.

    • Clarified the error message for hibernation or wake up operations on unsupported compute types.

  • Bug Fixes:

    • Fixed an issue where disk scaling settings did not take effect.

    • Fixed an issue where ECI instances were repeatedly woken up in rare scenarios.

    • Fixed an issue where the error cause was missing in some ECI instance wake-up failures.

This upgrade has no impact on your workloads.

April 2026

Version

Date

Description

Impact

v2.17.1

April 3, 2026

Fixed an issue from v2.16.3 and later that prevented the deletion of some ECI compute pods in a terminal phase (Failed or Succeeded).

This upgrade has no impact on your workloads.

v2.16.6

April 3, 2026

Fixed an issue from v2.16.3 and later that prevented the deletion of some ECI compute pods in a terminal phase (Failed or Succeeded).

This upgrade has no impact on your workloads.

March 2026

Version

Date

Description

Impact

v2.16.5

March 20, 2026

Improved pod update performance at scale.

No service impact.

February 2026

Version

Release date

Changes

Impact

v2.16.4

February 26, 2026

  • Improved pod creation performance at scale.

  • Improved pod-to-node binding performance at scale.

Non-disruptive update.

v2.16.3

February 5, 2026

  • Optimizations:

    • Improved component startup performance for large clusters.

    • Improved pod deletion performance for large clusters.

Non-disruptive update.

January 2026

Version

Update date

Changes

Impact

v2.16.2

January 28, 2026

  • Optimizations:

    • Optimized the retry mechanism for pod creation requests on virtual nodes when the quota is exceeded.

  • Bug Fixes:

    • Fixed an intermittent issue that caused pods on virtual nodes to get stuck in the Pending state due to unschedulable reserved nodes.

This upgrade has no impact on your workloads.

v2.16.1

January 27, 2026

  • Optimizations:

    • Optimized the component's internal HTTP connection pool management.

    • Further improved pod creation performance at scale from v2.16.0, reducing startup time.

  • Bug Fixes:

    • Fixed a rare, intermittent issue that caused pods to get stuck in the Pending state.

This upgrade has no impact on your workloads.

v2.16.0

January 6, 2026

  • New Features:

    • Enabled the pod capacity feature for virtual nodes by default.

    • Added support for the PodNetworking feature on ECI compute.

    • Added support for ECI compute to reference cross-namespace resources by using NodePublishSecretRef.

    • Added support for enabling the CSINode feature on reserved nodes.

    • Added support for reserved nodes to send event notifications during self-healing.

    • Added support for displaying the InterruptionHardwareFault condition on pods running on ECI GPU instances and reserved nodes.

  • Optimizations:

    • Improved component startup speed at scale.

    • Improved the success rate of instance cleanup in rare cases.

  • Bug Fixes:

    • Fixed an issue where reserved nodes could enter the NotReady state in specific scenarios.

    • Fixed an issue where pod conditions could be accidentally overwritten in extreme scenarios.

    • Fixed an authorization issue that could occur when using SidecarSet injection on ECI compute.

    • Fixed an issue where the Instance.ToBeReleased condition was not displayed correctly on ECI compute instances.

This upgrade has no impact on your workloads.

December 2025

Version

Date

Changes

Impact

v2.15.8

December 8, 2025

  • Optimizations:

    • Optimized local informer cache expiration handling.

    • Improved the processing speed of pods in a terminal phase.

  • Bug Fixes:

    • Fixed an issue where SidecarSet pod permissions were revoked.

    • Fixed an issue where conditions were overwritten in rare cases.

This update does not affect your services.

November 2025

Version

Release date

Description

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 update does not affect your workloads.

v2.15.6

November 5, 2025

  • Optimizations:

    • Optimized the resource normalization logic for ECI instances.

    • Improved the error message for pods on reserved nodes that use an unsupported ComputeClass.

  • Bug Fixes:

    • Fixed an issue where the local informer cache might not expire in time.

    • Fixed an issue where a pod's annotations could become inconsistent with its Spec after an in-place resource update.

This update does not affect your workloads.

October 2025

Version

Date

Changes

Impact

v2.15.5

October 20, 2025

  • Bug Fixes:

    • Further addressed an issue where the IP addresses of reserved nodes were missing in ACK clusters.

This upgrade has no impact on your services.

September 2025

Version

Date

Changes

Impact

v2.15.4

September 28, 2025

  • Optimizations:

    • Improved the error messages for scheduling DaemonSet Pods to virtual nodes.

    • Improved the recovery of GPU Pods from abnormal states in ACS clusters.

  • Bug Fixes:

    • Fixed an intermittent issue preventing ACS Pod status updates.

    • Fixed an issue where the IP address of a reserved node was missing in an ACS cluster.

    • Fixed intermittent crashes when using CNFS volumes.

This upgrade has no impact on your workloads.

v2.15.3

September 28, 2025

  • Bug Fixes:

    • Fixed an issue where Pods remained in the Pending state for long periods in rare cases.

    • Fixed an issue where Pods failed to reach a terminal state, such as Succeeded or Failed, in rare cases.

This upgrade has no impact on your workloads.

v2.15.2

September 22, 2025

  • Bug Fixes:

    • Fixed an issue in 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 cache data would expire due to delayed API server pushes.

This upgrade has no impact on your workloads.

v2.15.1

September 09, 2025

  • Fixed an issue where Pod creation failed when a vSwitch was specified via an annotation.

This upgrade has no impact on your workloads.

August 2025

Version

Release date

Description

Impact

v2.15.0

August 19, 2025

  • New features:

    • Introduced support for reserved nodes.

    • Introduced automatic protocol conversion from ECI to ACS.

  • Improvements:

    • Reduced memory usage by approximately 40%.

    • Improved the reliability of ACS pod creation.

    • Adjust the virtual node's ephemeral storage capacity (ephemeral-storage) to 60000 Ti to prevent scheduling failures.

  • Bug fixes:

    • Fixed a bug that caused ACS GPU pod scale-up to fail when a cloud disk was mounted.

    • Fixed failures during in-place container restarts.

    • Fixed an issue where the PodNetworking configuration does not take effect in an ACK cluster.

This upgrade has no impact on your workloads.

v2.12.0-acs.16

August 7, 2025

  1. Fixed an intermittent bug that caused pods to get stuck in the Pending state.

  2. Added status reporting for reserved nodes.

  3. Improved reporting for the Interruption.GPUShareReclaim condition.

This upgrade has no impact on your workloads.

May 2025

Version number

Release date

Description

Impact

v2.12.0-acs.14

May 12, 2025

  1. Fixed unexpected updates to reserved nodes.

  2. Fixed parsing failures caused by an invalid acs-profile format.

  3. Added support for exposing the Interruption.GPUShareReclaim Condition.

No impact on running workloads.

March 2025

Version number

Release date

Description

Impact

v2.12.0-acs.13

March 20, 2025

  1. Adds support for inventory-aware scheduling for GPU-accelerated pods.

This update does not affect your workloads.

February 2025

Version

Release date

Description

Impact

v2.12.0-acs.12

February 13, 2025

  1. Adjusted the ephemeral storage capacity on a virtual node to 60,000 TiB.

  2. Fixed a bug where ServiceUnavailable errors prevented retries during advanced scheduling.

  3. Optimized the interaction between a PVC and the csiprovisioner component in WaitForFirstConsumer mode.

No impact on your workloads.

January 2025

Version

Release date

Description

Impact

v2.12.0-acs.11

January 21, 2025

Optimized the advanced scheduling feature.

This upgrade does not affect your workloads.

v2.12.0-acs.10

January 17, 2025

  • Added support for exposing metrics for reserved nodes.

  • Optimized the retry mechanism.

  • Other feature optimizations.

This upgrade does not affect your workloads.

November 2024

Version

Date

Description

Impact

v2.12.0-acs.9

November 27, 2024

  • Supports scaling up cloud disks for GPU-accelerated Pods.

  • Supports stock awareness for general-purpose and performance-enhanced Pods.

  • Other feature improvements.

This upgrade has no impact on your workloads.

September 2024

Version number

Release date

Description

Impact

v2.12.0-acs.8

September 13, 2024

  • Upgraded the Kubernetes API dependency to v1.30.

  • Added the container_sockets and container_processes Prometheus metrics.

  • Enabled KubeSkoop to collect network metrics.

  • Enabled mounting disks to GPU-accelerated pods.

  • Enabled CPU hot scaling for general-purpose pods.

This upgrade has no impact on your workloads.

August 2024

Version

Release date

Description

Impact

v2.12.0-acs.7

August 21, 2024

Fixed an issue where an associated event was missing a UID.

This upgrade has no impact on your workloads.

v2.12.0-acs.6

August 2, 2024

General-purpose and performance Pods now support expanding cloud disk volumes.

This upgrade has no impact on your workloads.

July 2024

Version number

Release date

Description

Impact

v2.12.0-acs.5

July 25, 2024

Replaced role and role binding authentication with pod identity verification.

No impact on workloads.

June 2024

Version number

Release date

Description

Impact

v2.12.0-acs.4

June 21, 2024

  • Integrated with the scheduler to support Kubernetes-native scheduling.

  • Added support for creating Pods using a Capacity Reservation.

No impact on workloads.