ack-koordinator (FKA ack-slo-manager)-Container Service for Kubernetes(ACK)-阿里云帮助中心

ack-koordinator is a quality of service (QoS)-aware scheduling system. It uses features such as CPU Burst and dynamic resource overcommitment to optimize cluster resource use while ensuring the quality of service for high-priority applications, enhancing overall system stability. ack-koordinator also supports resource profiling and descheduling. This topic introduces the ack-koordinator component and its release notes.

Note

For the component's release and change history, see Release notes.

ack-koordinator

ack-koordinator is a Kubernetes-based QoS-aware scheduling system that improves cluster resource utilization and provides QoS guarantees for applications with different priorities. This prevents resource contention from affecting application performance. For example, ack-koordinator supports the colocation of service-oriented applications and batch processing tasks on the same node. It uses capabilities such as dynamic resource overcommitment and QoS-aware scheduling to improve resource utilization while ensuring QoS for service applications. This makes it suitable for batch processing, high-performance computing (HPC), AI tasks, and machine learning scenarios. ack-koordinator is the core component that enables QoS-aware scheduling in ACK. It provides features such as elastic resource limits, topology-aware scheduling, dynamic resource overcommitment, load-aware scheduling, descheduling, and resource profiling.

To get started, log on to the ACK console to install the ack-koordinator component in your ACK cluster. You can then enable and use its features with a ConfigMap or Pod annotations.

Architecture

ack-koordinator consists of central components and node-level components.

Koordinator Manager: A central component deployed as a Deployment. It consists of a primary and a standby instance to ensure high availability.
- SLO Controller: Manages resource overcommitment. It dynamically adjusts the overcommitted resources in the cluster based on the running status of nodes during colocation and manages differentiated SLO policies for each node.
- Recommender: Provides the resource profiling feature. It estimates the peak resource requirements for workloads to simplify the configuration of container resource specifications.
Koordinator Descheduler: A central component deployed as a Deployment that provides the descheduling feature.
Koordlet: A node-level component deployed as a DaemonSet. It supports dynamic resource overcommitment, load-aware scheduling, and QoS-aware scheduling in colocation scenarios.
Koordinator scheduling plug-in: The Koord-Scheduler module is not included when you install the ack-koordinator component. Instead, its scheduling capabilities are integrated as a plug-in into the ACK scheduler and installed by default.

Note

In an ACK Serverless cluster, ack-koordinator includes only the Koordinator Manager component and provides the resource profiling feature.

Versioning

Starting from v1.1.1-ack.1, the version number of ack-koordinator uses the x.y.z-ackn format.

x.y.z: Corresponds to the open source Koordinator version, meaning ack-koordinator supports all features of that version.
ackn: Represents feature enhancements and optimizations based on the open source version.

Key concepts

QoS

Kubernetes provides three Quality of Service (QoS) classes for Pods: Guaranteed, Burstable, and BestEffort. Koordinator is compatible with and enhances the Kubernetes QoS classes by introducing new QoS classes, such as Latency Sensitive (LS) and Best Effort (BE). Koordinator manages these new classes through the koordinator.sh/qosClass field. Koordinator QoS is designed to provide more fine-grained control and optimization of workload resource usage and scheduling in colocation scenarios to ensure the overall QoS of the system.

Descheduling

In Kubernetes, scheduling is the process of assigning pending Pods to nodes. Descheduling refers to the process of evicting Pods from suboptimal nodes and rescheduling them to more suitable ones. In scenarios where uneven cluster utilization creates hotspot nodes, or where changes in node attributes cause existing Pods to no longer match scheduling rules, you can use descheduling to optimize resource usage. This ensures that Pods run on the most suitable nodes, thereby guaranteeing high availability for the cluster and efficient operation for workloads.

Colocation

Colocation, also known as online and offline mixed deployment, is a strategy that runs workloads of different types and priorities on the same compute node. This helps improve resource utilization and reduce costs. This strategy fully utilizes compute and storage resources, reducing idle capacity to lower cluster costs. This approach is especially effective for workloads with fluctuating demand.

Supported features

The ack-koordinator component includes the features of the corresponding open source Koordinator version. During installation, only the feature-gates for common features are enabled by default. To use other features, you must manually enable their corresponding feature-gates for the ack-koordinator modules. For more information, see the Koordinator official documentation.

Type	Feature	Description	Consistency with open source
Load-aware scheduling		Schedules Pods to nodes with lower real-time loads to balance the cluster load and reduce the risk of node failure.	Yes
QoS-aware scheduling	CPU Burst	Improves application QoS by providing temporary, extra CPU resources during load spikes. This is achieved by automatically detecting CPU throttling and adaptively adjusting container parameters.	Yes
	CPU topology-aware scheduling	Pins a Pod to run on specific CPU cores of a node. This mitigates application performance degradation caused by CPU context switching and cross-NUMA memory access.	No
	Dynamic resource overcommitment	Quantifies allocated but unused CPU and memory resources by collecting real-time node load data, then provides these idle resources to BestEffort workloads while ensuring fairness.	Yes
	CPU Suppress	In dynamic resource overcommitment scenarios, this feature limits the CPU usage of BE Pods to protect the performance of LS Pods on the node.	Yes
	CPU QoS	Guarantees CPU resources for high-priority (LS) applications by preventing resource contention with lower-priority ones.	Yes
	Container memory QoS	Allows you to set QoS parameters for containers based on their priority. It prioritizes the performance of high-priority applications while ensuring memory resource fairness.	Yes
	L3 cache and memory bandwidth isolation	Isolates L3 cache and memory bandwidth usage between applications of different priorities to protect the performance of LS applications.	Yes
	Dynamic modification of Pod resource parameters	Lets you dynamically modify node-level isolation parameters (CPU, memory, and disk I/O) for a Pod or Deployment without a restart by changing its cgroup files.	No
	Data stream acceleration using hardware DSA	Leverages the Intel® Data Streaming Accelerator (DSA) to improve data processing efficiency for data-intensive workloads on a node. It also further enhances the acceleration of nearby memory access for containers.	No
	Container memory locality acceleration	Improves memory access performance for memory-intensive workloads by securely migrating memory from remote to local NUMA nodes for CPU-pinned applications. This process increases the local memory access hit rate.	No
Descheduling	Descheduling	Reschedules suboptimally placed Pods to healthier nodes to maintain cluster health, optimize resource usage, and improve workload QoS. This process is typically triggered by uneven resource utilization or high node load.	Yes
Descheduling	Load-aware hotspot descheduling	Automatically deschedules Pods from nodes that exceed a specified load watermark to prevent severe load imbalances.	Yes
GPU topology awareness	NUMA topology-aware scheduling	Optimizes performance by scheduling Pods to the most suitable NUMA node, which reduces cross-NUMA node access.	No
Resource profiling		Analyzes historical resource usage data to recommend resource specifications for containers, simplifying the process of configuring container `request` and `limit` values.	No

Installation and management

ack-koordinator is available on the Add-ons page in the ACK console. You can install, upgrade, and uninstall the component on the Add-ons page.

Prerequisites

You have an ACK cluster that runs Kubernetes 1.18 or later. For more information about how to upgrade a cluster, see Manually upgrade a cluster.
You have Helm v3.0 or later installed. To upgrade Helm, see [Component Upgrade] Announcement on Helm V2 Tiller Upgrade and How do I manually upgrade the version of Helm?.

Install and manage the component

You can install the ack-koordinator component on the Add-ons page. You can also return to this page to modify the component's parameters or upgrade it as needed.

For component versions earlier than v0.7 deployed from Marketplace, see Migrate ack-koordinator from Marketplace to add-ons to complete the migration.

Install the component: Follow the steps below to install the component and verify the deployment.
Modify component parameters: The system automatically redeploys ack-koordinator based on the new configuration.
Upgrade the component: If you have manually modified the deployed modules of ack-koordinator, such as Deployments or DaemonSets, the upgrade overwrites your custom configurations.

Log on to the ACK console. In the left navigation pane, click Clusters.
On the Clusters page, click the name of your cluster. In the left navigation pane, click Components and Add-ons .
On the Add-ons page, find ack-koordinator. In the ack-koordinator(ack-slo-manager) card, click Install.
In the Install ack-koordinator(ack-slo-manager) dialog box, modify the configuration parameters as needed and click OK.
(Optional) On the Cluster Management page, in the left-side navigation pane, select Applications to view the deployment status of ack-koordinator.
If the Status of ack-koordinator is Deployed, it indicates that the component has been successfully deployed.

Uninstall the component

Clean up ConfigMaps of existing nodes

The topology-aware CPU scheduling feature creates a topology ConfigMap in the kube-system namespace for each ACK node. Starting from v0.5.1, ack-koordinator automatically cleans up the ConfigMaps of removed nodes. However, when you uninstall ack-koordinator, it retains the ConfigMaps of existing nodes. These leftover ConfigMaps do not affect other features but occupy storage space. We recommend that you clean them up.

On the Add-ons page, find ack-koordinator and follow the on-screen instructions to uninstall the component.
Delete the topology ConfigMaps.
1. In the left-side navigation pane, choose Configurations > ConfigMaps. At the top of the page, select the kube-system namespace.
2. In the search box, enter -numa-info. From the list, find the ConfigMap that matches the ${NODENAME}-numa-info format. In the Actions column for the ConfigMap, click Delete and follow the on-screen instructions.

Clean up CRD objects

When you uninstall the ack-koordinator component, some CRD objects may remain. These leftover objects do not affect other features, but they occupy storage space. If you plan to reinstall ack-koordinator, these objects might interfere with the component's functionality. We recommend that you clean them up. The affected CRDs are listed below.

autoscaling.alibabacloud.com	slo.koordinator.sh
Expand `autoscaling.alibabacloud.com` to see version `v1alpha1` containing three CRD resource types: `RecommendationCheckpoint`, `Recommendation`, and `RecommendationProfile`.	Expand slo.koordinator.sh API Group to see version v1alpha1 containing two CRD resource types: NodeSLO and NodeMetric.

Billing

Installing or using the ack-koordinator component is free. However, you may incur costs in the following scenarios:

ack-koordinator is a non-managed component that consumes worker node resources after installation. You can configure resource requests for each module during installation.
By default, ack-koordinator exposes monitoring metrics for features such as resource profiling and fine-grained scheduling in Prometheus format. If you select the Enable Prometheus monitoring metrics for ack-koordinator option and use Managed Service for Prometheus, these metrics are billed as custom metrics. The fees vary based on factors such as cluster size and the number of applications. Before you enable this feature, we recommend reading Billing of Prometheus instances to understand the free tier and pricing for custom metrics. To monitor and manage your resource usage, use usage query.

Related information

Relationship between ack-koordinator and ack-slo-manager

ack-slo-manager is the predecessor of ack-koordinator and incubated the open source project Koordinator. As Koordinator matured, its technologies were incorporated back into ack-slo-manager. ack-koordinator offers the features of the open source Koordinator version and provides more capabilities than its predecessor. To access new features and bug fixes, upgrade to the latest component version by following the instructions in Migrate ack-koordinator from Marketplace to Add-ons.

Migrate ack-koordinator from Marketplace to Add-ons

If you deployed ack-koordinator from Marketplace (versions earlier than v0.7), you must uninstall and then reinstall it. Follow the instructions below to complete the migration.

If you modified the ConfigMap of the Marketplace ack-koordinator, back it up before you upgrade. If you have not modified the ConfigMap, skip to step 2 and upgrade the component directly.

Back up the ack-koordinator ConfigMap by using kubectl or the console.
kubectl
1. Run the following command to save the original configuration to the slo-config.yaml file. Replace the namespace (for example, kube-system) and name (for example, ack-slo-manager-config) with the actual values of your ConfigMap.
```
kubectl get cm -n kube-system ack-slo-manager-config -o yaml > slo-config.yaml
```
2. Run the vim slo-config.yaml command to edit the file. Change the namespace of the ConfigMap to kube-system, change the name field to ack-slo-config, and remove all annotations and labels from the ConfigMap to prevent them from being overwritten during the upgrade.
3. Run the following command to apply the modified configuration to the cluster.
```
kubectl apply -f slo-config.yaml 
```
Console
1. Record the key-value pairs of the original ConfigMap.
  1. In the navigation pane on the left, choose Configurations > ConfigMaps. At the top of the page, select the Namespace that you specified when you installed ack-koordinator from Marketplace. The default namespace is kube-system.
  2. In the Name search box, enter ack-slo-manager-config, click the name of the target ConfigMap, and record the key-value pairs in the ConfigMap.
2. Create a new ConfigMap by using the original ConfigMap's key-value pairs.
  1. In the navigation pane on the left, choose Configurations > ConfigMaps. At the top of the page, select All Namespaces.
  2. In the upper-right corner of the ConfigMaps page, click Create. Set ConfigMap Name to ack-slo-config and select the kube-system namespace. Click + Add, enter the recorded key-value pairs, and click Create.
In the navigation pane on the left, choose Applications > Helm. Find the ack-slo-manager component that was installed from Marketplace, and in the Actions column, click Delete to uninstall it.
On the Add-ons page, install the latest version of ack-koordinator. For more information, see Install and manage components.
Important
If you modified the ConfigMap for the ack-koordinator component from Marketplace, you must enter the name of the backup ConfigMap that you created in Step 1 (for example, ack-slo-config) into the corresponding parameter field in the ack-koordinator Parameters dialog box.

Migrate from resource-controller to ack-koordinator

The resource-controller component has been deprecated. ack-koordinator now supports all features of resource-controller, such as topology-aware CPU scheduling and dynamically modifying the resource parameters of a pod. If your cluster uses resource-controller, follow these steps to migrate from resource-controller to ack-koordinator.

Upgrade resource-controller to the latest version.
1. Log on to the ACK console. In the left navigation pane, click Clusters.
2. On the Clusters page, click the name of your cluster. In the left navigation pane, click Components and Add-ons .
3. On the Add-ons page, find resource-controller and follow the on-screen instructions to upgrade the component.
Install and configure ack-koordinator.
1. On the Add-ons page, find ack-koordinator and follow the on-screen instructions to install it.
2. In the Installack-koordinator(ack-slo-manager) dialog box, configure the agentFeatures parameter, which is the feature-gate switch for Koordlet, and other parameters as needed. Then, click OK.
  1. Check whether your cluster uses the CPU Burst feature described in Dynamically modify the resource parameters of a pod. This feature modifies the cpu.cfs_quota_us cgroups file for a specific container by creating a CRD or adding a pod annotation. If you use this feature, go to step ii. Otherwise, skip to step c.
  2. Run the following command to get the current feature-gate configuration from the DaemonSet YAML file of ack-koordlet.
```
kubectl get daemonset -n kube-system ack-koordlet -o yaml |grep feature-gates
- --feature-gates=AllAlpha=false,AllBeta=false,...,CPUBurst=true,....
```
  3. Modify the feature-gate configuration for ack-koordlet to disable the CPU Burst performance optimization policy by setting CPUBurst=false. Use commas (,) to separate parameters. Keep other settings unchanged.
    Disabling the policy also disables the CPU Burst mechanism for all containers in the cluster. This prevents two modules from simultaneously modifying the cpu.cfs_quota_us cgroups file.
```
AllAlpha=false,AllBeta=false,...,CPUBurst=false,....
```
    To enable elastic CPU resources for containers, use the CPU Burst performance optimization policy to automatically adjust the CPU elasticity of pods. For more information, see Enable the CPU Burst performance optimization policy.
3. In the navigation pane on the left, choose Applications > Helm to view the deployment status of ack-koordinator.
  If the Status for ack-koordinator is Deployed, the component is successfully deployed.
On the Add-ons page, find resource-controller and follow the on-screen instructions to uninstall it.

FAQ

Component installation error: no matches for kind "ServiceMonitor" in version "monitoring.coreos.com/v1" ensure CRDs are installed first

Managed Service for Prometheus is not installed in the cluster. Install it by following the instructions in Connect and configure Managed Service for Prometheus, or deselect Enable Prometheus Metrics for ACK-Koordinator when you install ack-koordinator on the Add-ons page.

Component installation error: task install-addons-xxx timeout, error install addons map[ack-slo-manager:Can't install release with errors: ... function "lookup" not defined

Upgrade Helm to v3.0 or later. See Component upgrade: Helm V2 Tiller upgrade announcement for more information.

Release notes

July 2026

Version	Image address	Date	Changes	Impact
v1.8.0-ack1.25	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.8.0-ack1.25-1ae0ed2-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.8.0-ack1.25-1ae0ed2-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.8.0-ack1.25-1ae0ed2-aliyun	July 29, 2026	Refactored internal interfaces for optimization.	None

January 2026

Version	Image address	Date	Changes	Impact
v1.6.1-ack1.23	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.6.1-ack1.23-dc52dce-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.6.1-ack1.23-dc52dce-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.6.1-ack1.23-dc52dce-aliyun	January 30, 2026	Fixed a logging issue in the descheduler. You can now configure the `checkNewNodeOverUtilization` switch in the LowNodeLoad plugin.	None

November 2025

Version	Image address	Release date	Description	Impact
v1.6.1-ack1.21	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.6.1-ack1.21-561945f-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.6.1-ack1.21-561945f-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.6.1-ack1.21-561945f-aliyun	November 06, 2025	Fixed a compatibility issue in the ack-koordlet module with containerd v2.1.1.	None

August 2025

Version	Image address	Release date	Description	Impact
v1.6.1-ack1.19	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.6.1-ack1.19-a1cdc4c-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.6.1-ack1.19-a1cdc4c-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.6.1-ack1.19-a1cdc4c-aliyun	August 08, 2025	Fixed a regression from v1.6.1-ack1.16 by correcting the default node utilization statistics window to 5 minutes.	None

July 2025

Version	Image address	Release date	Description	Impact
v1.6.1-ack1.18	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.6.1-ack1.18-2f130b1-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.6.1-ack1.18-2f130b1-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.6.1-ack1.18-2f130b1-aliyun	July 30, 2025	Optimized the startup speed and stability of the ack-koord-manager module. Corrected the display of colocated resource request metrics on the colocation monitoring dashboard.	None
v1.6.1-ack1.17	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.6.1-ack1.17-1c86593-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.6.1-ack1.17-1c86593-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.6.1-ack1.17-1c86593-aliyun	July 14, 2025	Fixed a regression from v1.6.1-ack1.16 that prevented normal operation on Arm nodes.	None
v1.6.1-ack1.16	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.6.1-ack1.16-1c86593-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.6.1-ack1.16-1c86593-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.6.1-ack1.16-1c86593-aliyun	July 04, 2025	Optimized the descheduling feature for multi-template configurations and fixed an issue with eviction throttling in such configurations. Refactored internal interfaces for optimization.	None

September 2024

Version	Image address	Release date	Description	Impact
v1.5.0-ack1.14	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.5.0-ack1.14-4225002-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.5.0-ack1.14-4225002-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.5.0-ack1.14-4225002-aliyun	September 12, 2024	Optimized the calculation logic for the descheduling `maxUnavailablePerWorkload` parameter. Fixed the dynamic adjustment of pod resource parameters using cgroups. Renamed metrics related to resource profiling. Optimized internal APIs.	None

July 2024

Version	Image address	Release date	Description	Impact
v1.5.0-ack1.12	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.5.0-ack1.12-a096ae3-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.5.0-ack1.12-a096ae3-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.5.0-ack1.12-a096ae3-aliyun	July 29, 2024	Optimized internal interfaces	None

January 2024

Version	Image address	Release date	Description	Impact
v1.3.0-ack1.8	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.3.0-ack1.8-94730e0-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.3.0-ack1.8-94730e0-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.3.0-ack1.8-94730e0-aliyun	2024-01-24	Adds support for installation on ACK Lingjun clusters. Maintains feature parity with v1.3.0-ack1.6.	None

December 2023

Version	Image address	Release date	Description	Impact
v1.3.0-ack1.7	registry.{region}.aliyuncs.com/acs/koord-manager:v1.3.0-ack1.7-3500ce9-aliyun registry.{region}.aliyuncs.com/acs/koordlet:v1.3.0-ack1.7-3500ce9-aliyun registry.{region}.aliyuncs.com/acs/koord-descheduler:v1.3.0-ack1.7-3500ce9-aliyun	December 21, 2023	Enables installation on ACK Edge clusters. Installs only on cloud worker nodes, not on edge nodes. No feature changes from v1.3.0-ack1.6.	None

October 2023

Version	Image	Release date	Description	Impact
v1.3.0-ack1.6	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.3.0-ack1.6-3500ce9-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.3.0-ack1.6-3500ce9-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.3.0-ack1.6-3500ce9-aliyun	October 19, 2023	Optimized internal interfaces.	None

June 2023

Version	Image address	Release date	Description	Impact
v1.2.0-ack1.3	registry-cn-zhangjiakou.ack.aliyuncs.com/acs/koord-manager:v1.2.0-ack1.3-89a9730-aliyun registry-cn-zhangjiakou.ack.aliyuncs.com/acs/koordlet:v1.2.0-ack1.3-89a9730-aliyun registry-cn-zhangjiakou.ack.aliyuncs.com/acs/koord-descheduler:v1.2.0-ack1.3-89a9730-aliyun	2023-06-09	Optimized internal interfaces.	None

April 2023

Version	Image address	Release date	Description	Impact
v1.2.0-ack1.2	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.2.0-ack1.2-b675c9a8-aliyun registry.{REGION}.aliyuncs.com/acs/koordlet:v1.2.0-ack1.2-b675c9a8-aliyun registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.2.0-ack1.2-b675c9a8-aliyun	April 25, 2023	Supports accelerating nearby memory access for containers. Supports data stream acceleration using hardware DSA. Optimizes internal APIs.	None

March 2023

Version	Image address	Release date	Description	Impact
v1.1.1-ack.2	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.1.1-ack.2 registry.{REGION}.aliyuncs.com/acs/koordlet:v1.1.1-ack.2 registry.{REGION}.aliyuncs.com/acs/koord-descheduler:v1.1.1-ack.2	March 23, 2023	Optimized internal interfaces.	None

January 2023

Version	Image	Release date	Description	Impact
v1.1.1-ack.1	registry.{REGION}.aliyuncs.com/acs/koord-manager:v1.1.1-ack.1 registry.{REGION}.aliyuncs.com/acs/koordlet:v1.1.1-ack.1	January 11, 2023	Component renamed to ack-koordinator. Optimized internal APIs.	None

November 2022

Version	Image address	Release date	Description	Impact
v0.8.0	registry.{REGION}.aliyuncs.com/acs/koord-manager:v0.8.0 registry.{REGION}.aliyuncs.com/acs/koordlet:v0.8.0	November 17, 2022	Updated load-aware pod scheduling to support Kubernetes 1.22. Optimized internal APIs.	After upgrading the component, you must upgrade your ACK cluster to version 1.22.15-ack-2.0 to use load-aware pod scheduling. This upgrade does not affect other features.

September 2022

Version	Image address	Release date	Description	Impact
v0.7.2	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.7.2	September 16, 2022	Fixed an issue introduced in v0.7.1 where topology-aware scheduling failed to apply to pods.	None
v0.7.1	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.7.1	September 2, 2022	Fixed a CPU throttling issue with topology-aware scheduling in CentOS 3 kernel environments. Added support for enabling and disabling Managed Service for Prometheus. Added resource profiling features to the console. Removed the option to install from the Marketplace.	None

August 2022

Version	Image	Release date	Description	Impact
v0.7.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.7.0	August 8, 2022	ack-slo-manager installation has moved from Marketplace to Add-ons.	None

July 2022

Version	Image address	Release date	Description	Impact
v0.6.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.6.0	July 26, 2022	Optimized internal APIs and simplified component configuration.	None

June 2022

Version	Image	Date	Description	Impact
v0.5.2	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.5.2	June 14, 2022	Optimized internal interfaces. Optimized CPU QoS for containers. Added an option to disable automatic ack-slo-manager namespace creation.	None
v0.5.1	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.5.1	June 2, 2022	Added support for CPU QoS for containers. Optimized internal interfaces and automated the cleanup of topology ConfigMaps for removed nodes.	None

April 2022

Version	Image address	Release date	Description	Impact
v0.5.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.5.0	April 29, 2022	Supports dynamically modifying pod resource parameters. Supports Arm architecture environments.	None
v0.4.1	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.4.1	April 14, 2022	Supports CPU Suppress. Supports resource isolation based on the L3 cache and Memory Bandwidth Allocation (MBA). Optimized internal APIs.	None
v0.4.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.4.0	April 11, 2022	Optimized memory consumption of slo-agent.	None

February 2022

Version	Image address	Release date	Description	Impact
v0.3.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.3.0	February 25, 2022	Added support for resource profiling. Added support for container memory QoS. Added support for dynamic resource overcommitment. Optimized internal APIs.	None

December 2021

Version	Image address	Release date	Description	Impact
v0.2.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.2.0	2021-12-10	Introduced the CPU burst performance optimization policy. Introduced topology-aware CPU scheduling.	None

September 2021

Version	Image address	Release date	Description	Impact
v0.1.1	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.1.1-c2ccefa	September 02, 2021	Optimized internal interfaces.	None

July 2021

Version	Image address	Release date	Description	Impact
v0.1.0	registry.{REGION}.aliyuncs.com/acs/ack-slo-manager:v0.1.0-09766de	July 8, 2021	Now supports load-aware pod scheduling.	None