karmada-io
/
karmada
mirror of https://github.com/karmada-io/karmada.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
karmada/docs/CHANGELOG/CHANGELOG-1.7.md

13 KiB
Raw Blame History

Table of Contents generated with DocToc

v1.7.0

Downloads for v1.7.0

Download v1.7.0 in the v1.7.0 release page.

What's New

CronFederatedHPA

Introduced CronFederatedHPA API, which represents a collection of repeating schedules to scale the replica number of a specific workload. It is used for regular auto scaling scenarios, and it can scale any workloads that have a scale subresource or FederatedHPA.

See CronFederated HPA documents for more details.

(Feature contributors: @jwcesign, @chaunceyjiang, @RainbowMango)

MultiClusterService

Introduced MultiClusterService API, which controls the exposure of services to multiple external clusters and also enables service discovery between clusters.

See Expose and discover multi-cluster services for more details.

(Feature contributors: @XiShanYongYe-Chang, @chaunceyjiang, @yike21)

PropagationPolicy Preemption

The PropagationPolicy/ClusterPropagationPolicy is now able to preempt resources from another PropagationPolicy/ClusterPropagationPolicy as per priority by declaring the preemption behavior. Preemption will be disabled by default for backward compatibility.

This feature is now controlled by feature gates PropagationPolicyPreemption with the alpha state (disabled by default). See PropagationPolicy Priority and Preemption for more details.

(Feature contributors: @Poor12, @whitewindmills, @jwcesign, @RainbowMango)

Migrating resources in batches

Legacy cluster resources can now be migrated to Karmada in batches by PropagationPolicy or ClusterPropagationPolicy. By specifying the ConflictResolution in the PropagationPolicy/ClusterPropagationPolicy, workloads can be migrated to Karmada smoothly without container termination or restarts.

See Migrate In Batch for more details.

(Feature contributors: @chaosi-zju, @RainbowMango)

FederatedHPA

FederatedHPA is now able to scale replicas based on custom metrics in addition to CPU and memory.

See FederatedHPA scales with custom metrics for more details.

(Feature contributor: @chaunceyjiang)

Other Notable Changes

API Changes

  • Added print columns for FederatedHPA API including reference, minpods, maxpods, and replicas. (#3622, @Poor12)
  • Added CronFederatedHPA API based on the proposal. (#3692, @RainbowMango)
  • Introduced Preemption to both PropagationPolicy and ClusterPropagationPolicy to declare the behaviors of preemption. (#3788, @RainbowMango)
  • Introduced ConflictResolution to both PropagationPolicy and ClusterPropagationPolicy to declare how potential conflicts should be handled. (#3801, @RainbowMango)
  • Introduced a new field Zones to represent multiple zones of a member cluster. (#3933, @RainbowMango)

Deprecation

  • karmada-controller-manager: Removed hpa controller in favor of FederatedHPA. (#3852, @jwcesign)
  • The Zone(.spec.zone) was never used and now has been deprecated in favor of the newly introduced Zones. (#3933, @RainbowMango)
  • karmadactl: Deprecated --cluster-zone flag which will be removed in future releases and introduced --cluster-zones flag. (#3995, @whitewindmills)

Bug Fixes

  • karmada-metrics-adapter: Fixed the issue that when different clusters have the same pod name, the annotations of different metrics have the same value. (#3647, @chaunceyjiang)
  • karmada-controller-manager: Fixed panic when printing log in the case that lastSuccessfulTime of cronjob is nil. (#3683, @chaunceyjiang)
  • karmada-controller-manager: Fixed the issue that the Applied condition of ResourceBinding is always true. (#3709, @chaunceyjiang)
  • karmada-controller-manager: Fixed a boundary case where the propagationpolicy.karmada.io/name in the label of the resource template would not be removed after deleting PP. (#3848, @chaunceyjiang)
  • karmada-controller-manager: Fixed the issue that dependent resources are created and deleted repeatedly when the dependent resource has a status field. (#3868, @chaunceyjiang)
  • karmada-controller-manager: Avoid updating directly cached resource templates. (#3879, @whitewindmills)
  • kamrada-controller-manager: Fixed the issue that federated-HPA plain metric calculation is incorrect when usageRatio == 1.0, keep same with resource replicas. (#3922, @zach593)
  • karmada-webhook: When application failover is enabled, users are prevented from setting propagateDeps to false. (#3739, @chaunceyjiang)
  • karmada-search: Fixed a panic due to concurrent mutating objects in the informer cache. (#3966, @ikaven1024)
  • Fixed the inability to sync list issues in the case that the client disconnects from the member cluster. The fixes apply to the following components:
    • karmada-controller-manager
    • karmada-agent
    • karmada-descheduler
    • karmada-scheduler
    • karmada-metrics-adapter
    • karmada-search (#3908, @WulixuanS)

Security

  • The base image alpine now has been promoted from alpine:3.18.2 to alpine:3.18.3. (#3942, @Rajan-226)
  • Security: Sets an upper bound for all components on the lifetime of idle keep-alive connections and time to read the headers of incoming requests. (#3951, @zishen)

Features & Enhancements

  • karmadactl: The --wait-component-ready-timeout flag has been introduced in the init command to specify the component installation timeout. (3665, @helen-frank)
  • karmadactl: Added karmada-metrics-adapter to addons. ([#3717](karmadactl: Added karmada-metrics-adapter to addons.), @chaunceyjiang)
  • karmadactl: Introduced top command. (#3593, @chaunceyjiang)
  • karmadactl join/register, karmada-agent: Added labels on the namespace created by Karmada. (#3839, @zhy76)
  • karmadactl: Granted full permissions of Karmada resources to admin during deployment of Karmada with init. (#3937, @zhy76)
  • karmada-controller-manager: Supported aggregating the status of a pod's initContainer. (#3574, @chaunceyjiang)
  • karmada-controller-manager: Supported modification synchronization of custom resources as dependencies. (#3614, @chaunceyjiang)
  • karmada-controller-manager: Implemented proxy header of cluster APIs. (#3631, @Poor12)
  • karmada-controller-manager: The --cluster-cache-sync-timeout flag is now used to specify the sync timeout of the control plane cache in addition to the member cluster's cache. The default value has been increased to 2 minutes. (#3874, @lxtywypc)
  • karmada-controller-manger: Added Ingresses to the default dependencinterpreter. (#3885, @chaunceyjiang)
  • karmada-controller-manager: Introduced a LabelSelector field to DependentObjectReference. (#3811, @chaunceyjiang)
  • karmada-controller-manager: Added the pod replica interpreter by default. (#3876, @whitewindmills)
  • karmada-controller-manager: Added dependencies of ServiceImport to the default interpreter. (#3939, @chaunceyjiang)
  • karmada-controller-manager: Included the UID of the owner resource in labels and provided the details in annotations. If users are using related labels as label selectors, they should switch to using the UID as the label selector. (#4007, @jwcesign)
  • karmada-scheduler: Introduced new scheduling condition reasons: NoClusterFit, SchedulerError, Unschedulable, Success. (#3741, @whitewindmills)
  • karmada-operator: Supported disabling karmada cascading deletion. (#3577, @calvin0327)
  • karmada-operator: Allowed installing the karmada-metrics-adapter addon. (#3732, @calvin0327)
  • karmada-webhook: Allowed custom metrics configuration of FederatedHPA. (#3826, @jwcesign)

Other

Dependencies

  • The base image alpine now has been promoted from alpine:3.17.1 to alpine:3.18.2. (#3671, @yanggangtony)
  • Karmada is now built with Kubernetes v1.27.3 dependencies. (#3730, @RainbowMango)
  • Karmada(v1.7) is now built with Go 1.20.6. (#3865, @parthn2)

Helm Charts

  • Fixed the issue that karmada-search no ETCD secret volume mount when using external ETCD. (#3777, @my-git9)
  • Granted full permissions of Karmada resources to admin. (#3957, @Vacant2333)
  • Now able to render PodDisruptionBudget for resources. (#3955, @a7i)

Instrumentation

  • Karmada images now are signed with cosign. (3434, @liangyuanpeng)
  • Removed specific labels from the following metrics of karmada-controller-manager to reduce the metrics count:
    • resource_match_policy_duration_seconds: removed apiVersion/kind/name/namespace.
    • resource_apply_policy_duration_seconds: removed apiVersion/kind/name/namespace.
    • policy_apply_attempts_total: removed namespace/name.
    • binding_sync_work_duration_seconds: removed namespace/name.
    • work_sync_workload_duration_seconds: removed namespace/name. (#3795, @jwcesign)
  • Introduced the TaintClusterSucceed event to Cluster object and merged TaintClusterByConditionFailed and RemoveTargetClusterFailed to TaintClusterFailed. (#2736, @Poor12)
  • Introduced cronfederatedhpa_process_duration_seconds and cronfederatedhpa_rule_process_duration_seconds metrics for CronFederatedHPA. (#3979, @whitewindmills)
  • Introduced federatedhpa_process_duration_seconds and federatedhpa_pull_metrics_duration_seconds for FederatedHPA, which will be emitted by karmada-controller-manager. (#3972, @zhy76)