Time | Version | Update |
2026-08-26 | v1.36.2-tke.1 | Replace the kube-proxy base image to support nftables mode. (kube-proxy) Synchronize the community code of version 1.36.2. (all) |
Time | Version | Update |
2026-07-22 | v1.34.1-tke.8 | Replace the kube-proxy base image to support iptables backend self-adaptation and ipset version compatibility, resolving abnormal issues caused by version mismatches. (Dockerfile) Fixed the issue where upgrading a tke cluster from 1.32 to 1.34 triggered ghost rollouts of existing StatefulSets. (kube-controller-manager) Fixed the issue where kube-controller-manager panicked when processing Service watch events because spec.ipFamilies was empty. (kube-controller-manager) |
2026-06-29 | v1.34.1-tke.6 | The crane-scheduler ClusterNodeResourcePolicy supports scaling up extended resources such as hugepages. (kube-scheduler) kubelet supports disabling cadvisor's reading of proc FD via a feature-gate. kube-controller-manager supports progressive creation of DaemonSets. After a restart, kubelet no longer temporarily changes the Ready status of existing Pods/containers due to the initialization of readiness/startup probe caches. (kubelet) |
2026-03-13 | v1.34.1-tke.5 | Fixed the memory leak issue in the scheduler. (kube-scheduler) |
2026-01-27 | v1.34.1-tke.4 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) |
2026-01-06 | v1.34.1-tke.3 | Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-17 | v1.34.1-tke.2 | Fixed the issue where Pods encounter an UnexpectedAdmissionError when nodes in vpc-cni mode restart. (kubelet) |
2025-11-19 | v1.34.1-tke.1 | Fixed the issue that when Pods were deployed on nodes in a cluster to mount PVCs, PVC monitoring data was not synchronized. (kubelet) |
Time | Version | Update |
2026-07-22 | v1.32.2-tke.14 | Replace the kube-proxy base image to support iptables backend self-adaptation and ipset version compatibility, resolving abnormal issues caused by version mismatches. (Dockerfile) |
2026-06-29 | v1.32.2-tke.12 | kubelet supports disabling cadvisor's reading of proc FD via a feature-gate. kube-controller-manager supports progressive creation of DaemonSets. After a restart, kubelet no longer temporarily changes the Ready status of existing Pods/containers due to the initialization of readiness/startup probe caches. (kubelet) |
2026-03-13 | v1.32.2-tke.11 | Fixed the memory leak issue in the scheduler. (kube-scheduler) |
2026-01-27 | v1.32.2-tke.10 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) |
2026-01-06 | v1.32.2-tke.9 | Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-17 | v1.32.2-tke.8 | Fixed the issue where Pods encounter an UnexpectedAdmissionError when nodes in vpc-cni mode restart. (kubelet) Fixed the issue of collecting containerd disk metrics. (kubelet) |
2025-11-19 | v1.32.2-tke.7 | Fixed the issue of collecting container monitoring information. (kubelet) Optimized the processing of pagination requests in the etcd overload protection feature. (kube-apiserver) |
2025-09-08 | v1.32.2-tke.6 | Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) |
2025-08-07 | v1.32.2-tke.5 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.32.2-tke.4 | Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-07-09 | v1.32.2-tke.3 | Fixed the monitoring and reporting issues in the etcd overload protection feature. (kube-apiserver) Enabled kubelet to support Pod deletion protection. (kubelet) |
2025-05-15 | v1.32.2-tke.2 | Optimized the reverse probing node of kcm to parallel probing. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.32.2-tke.1 | Added support for specifying Pods during scale-in. (kube-controller-manager) Implemented etcd overload protection. (kube-apiserver) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Optimized the reverse probing node of kcm. (kube-controller-manager) Optimized the performance of HPA. (kube-controller-manager) Changed the default validity period of apiserver self-signed certificates to 100 years. (kube-apiserver) Supported precise rate limiting for high-overhead List requests, and one-click circuit breaking for system and business component List requests. (kube-apiserver) Supported independent apiserver. (kube-apiserver) Synchronize the community code of version 1.32.4. (all) |
Time | Version | Update |
2026-07-22 | v1.30.0-tke.22 | Replace the kube-proxy base image to support iptables backend self-adaptation and ipset version compatibility, resolving abnormal issues caused by version mismatches. (Dockerfile) Fixed the issue where kubelet did not obtain the imageRecordsLock when removing images. (kubelet) |
2026-06-29 | v1.30.0-tke.20 | Fixed the resize status error in kubelet caused by inconsistent scaling during Pod memory resource comparison. (kubelet) kube-controller-manager supports progressive creation of DaemonSets. |
2026-03-13 | v1.30.0-tke.19 | Fixed the memory leak issue in the scheduler. (kube-scheduler) |
2026-01-27 | v1.30.0-tke.18 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) |
2026-01-06 | v1.30.0-tke.17 | Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-17 | v1.30.0-tke.16 | Fixed the issue where Pods encounter an UnexpectedAdmissionError when nodes in vpc-cni mode restart. (kubelet) During eviction, kube-controller-manager forcibly deletes pods on the failed node that have the annotation force-delete-when-node-unreachable=true. |
2025-11-19 | v1.30.0-tke.15 | Optimized the processing of pagination requests in the etcd overload protection feature. (kube-apiserver) |
2025-09-08 | v1.30.0-tke.14 | Fixed the issue of missing hostnames in node addresses. (kubelet) Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) Supported disabling cadvisor reading proc fd through feature-gate. (kubelet) Ported community PRs 127396 and 128219 to address the issue where kubelet reported a large number of vol_data.json: no such file or directory errors. (kubelet) |
2025-08-07 | v1.30.0-tke.13 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.30.0-tke.12 | Changed the validity period of cluster certificates generated by kubeadm from 1 year to 100 years. (kubeadm) Enabled kubelet to support Pod deletion protection. (kubelet) Supported in-place Pod specification adjustment. (apiserver/scheduler/kubelet) Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-05-15 | v1.30.0-tke.11 | Optimized the reverse probing node of kcm. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.30.0-tke.10 | PR125675: Fixed the update inconsistency issue with endpoints. (kube-controller-manager) Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) Fixed the cache issue of monthly subscription super nodes in the scheduler. (kube-scheduler) Optimized the reverse probing node of kcm to parallel probing. (kube-controller-manager) Supported independent apiserver. (kube-apiserver) Synchronized community code to version 1.30.12. (all) |
2025-04-01 | v1.30.0-tke.9 | Fixed the issue of illegal regular expression matching. (kube-apiserver) |
2025-03-25 | v1.30.0-tke.8 | PR3672: kubelet fixed the cadvisor cri client timeout issue. (kubelet) Supported precise speed limit for high-overhead list requests, one-click circuit-breaking systems, and business components such as List. (kube-apiserver) Supported configuring etcd overload protection via configmap. (kube-apiserver) Enabled kube-proxy to support the sh+port scheduling method in ipvs mode. (kube-proxy) |
2025-03-06 | v1.30.0-tke.7 | Fixed the memory loss issue and data forwarding plane issue of kube-proxy. (kube-proxy) |
2025-02-25 | v1.30.0-tke.6 | PR126543: Fixed the issue where init containers were stuck in the "created" state. (kubelet) |
2025-01-09 | v1.30.0-tke.5 | Ported community PR124223 to fix the issue where StorageObjectCountTracker was nil. (kube-apiserver) Optimized resource awareness policies. (kube-scheduler) Mitigated apiserver cache breakdown. (kube-apiserver) |
2024-12-20 | v1.30.0-tke.4 | Optimized the performance of HPA. (kube-controller-manager) Added the qcloud csi migration capability to ensure that in-tree pv Pods in clusters upgraded from early versions can be properly terminated and recreated. (kube-controller-manager kube-scheduler kubelet) Supported scale-out in a specified sequence by group. (kube-scheduler) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-09-24 | v1.30.0-tke.3 | Optimized the startup of the kube-proxy ipvs mode. (kube-proxy) |
2024-09-20 | v1.30.0-tke.2 | Fixed the issue where a get pod error was reported after TKE Serverless Pods were injected into ds. (kube-apiserver) Merged PR125039 to fix the scheduler's modulo-by-zero issue. (kube-scheduler) Optimized scheduler performance. (kube-scheduler) |
2024-07-24 | v1.30.0-tke.1 | Added support for specifying Pods during scale-in. (kube-controller-manager) Supported cgroupfs. (kubelet) Implemented etcd overload protection. (kube-apiserver) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Added annotation support for mounting cgroupfs to Pods. (kubelet). Supported the native node scaling feature. (kube-scheduler) Correctly displayed disk and network metrics when the runtime is containerd. (kubelet) Super nodes support DaemonSets. (kube-apiserver/kube-controller-manager) Optimized HPA for super nodes. (kube-controller-manager). Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Changed the validity period of the apiserver's self-signed certificate from 1 year to 5 years. (kube-apiserver) Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). |
Date | Version | Update |
2026-07-22 | v1.28.3-tke.26 | Replace the kube-proxy base image to support iptables backend self-adaptation and ipset version compatibility, resolving abnormal issues caused by version mismatches. (Dockerfile) |
2026-06-29 | v1.28.3-tke.24 | Fixed the resize status error in kubelet caused by inconsistent scaling during Pod memory resource comparison. (kubelet) kube-controller-manager supports progressive creation of DaemonSets. |
2026-01-27 | v1.28.3-tke.23 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) |
2026-01-06 | v1.28.3-tke.22 | Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-17 | v1.28.3-tke.21 | Fixed the issue where Pods encounter an UnexpectedAdmissionError when nodes in vpc-cni mode restart. (kubelet) During eviction, kube-controller-manager forcibly deletes pods on the failed node that have the annotation force-delete-when-node-unreachable=true. |
2025-11-19 | v1.28.3-tke.20 | Optimized the processing of pagination requests in the etcd overload protection feature. (kube-apiserver) |
2025-09-08 | v1.28.3-tke.19 | Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) Supported disabling cadvisor reading proc fd through feature-gate. (kubelet) Ported community PRs 127396 and 128219 to address the issue where kubelet reported a large number of vol_data.json: no such file or directory errors. (kubelet) |
2025-08-07 | v1.28.3-tke.18 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.28.3-tke.17 | Supported in-place Pod specification adjustment. (apiserver/scheduler/kubelet) Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-05-15 | v1.28.3-tke.16 | Optimized the reverse probing node of kcm. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.28.3-tke.15 | PR125675: Fixed the update inconsistency issue with endpoints. (kube-controller-manager) Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) Fixed the cache issue of monthly subscription super nodes in the scheduler. (kube-scheduler) Optimized the reverse probing node of kcm to parallel probing. (kube-controller-manager) Supported independent apiserver. (kube-apiserver) Synchronized community code to version 1.28.15. (all) |
2025-04-01 | v1.28.3-tke.14 | Fixed the issue of illegal regular expression matching. (kube-apiserver) |
2025-03-25 | v1.28.3-tke.13 | PR3672: kubelet fixed the cadvisor cri client timeout issue. (kubelet) Supported precise speed limit for high-overhead list requests, one-click circuit-breaking systems, and business components such as List. (kube-apiserver) Supported configuring etcd overload protection via configmap. (kube-apiserver) |
2025-03-06 | v1.28.3-tke.12 | Fixed the memory loss issue and data forwarding plane issue of kube-proxy. (kube-proxy) |
2025-02-25 | v1.28.3-tke.11 | Optimized the scoring of scheduling policies. (kube-scheduler) |
2025-01-09 | v1.28.3-tke.10 | Ported community PR124223 to fix the issue where StorageObjectCountTracker was nil. (kube-apiserver) Optimized resource awareness policies. (kube-scheduler) Mitigated apiserver cache breakdown. (kube-apiserver) |
2024-12-20 | v1.28.3-tke.9 | Supported scale-out in a specified sequence by group. (kube-scheduler) The SchedulerQueueingHints feature gate had many bugs and was disabled by default in Community Edition. (kube-scheduler) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-11-27 | v1.28.3-tke.8 | Fixed the issue where static PV Pods could not be scheduled when the cluster had chc nodes. (kube-scheduler) |
2024-09-24 | v1.28.3-tke.7 | Optimized the startup of the kube-proxy ipvs mode. (kube-proxy) |
2024-09-20 | v1.28.3-tke.6 | Merged PR123570 to fix the duplicate path issue with apiservice. (kube-apiserver) Fixed the issue where a get pod error was reported after TKE Serverless Pods were injected into ds. (kube-apiserver) Optimized scheduler performance. (kube-scheduler) Added the qcloud CSI migration capability to ensure normal termination and recreation of intree PV Pods after version upgrades. (kube-controller-manager/kube-scheduler/kubelet) Supported in-place Pod specification upgrades. (kube-apiserver/kube-scheduler/kubelet) Enabled kube-proxy to support the sh+port scheduling method in ipvs mode. (kube-proxy) |
2024-07-09 | v1.28.3-tke.5 | Enabled kubelet to support Pod deletion protection. (kubelet) Optimized scheduling logic. (kube-scheduler) |
2024-05-24 | v1.28.3-tke.4 | Implemented etcd overload protection. (kube-apiserver) Fixed the issue where super nodes did not assign the CIDR. (kube-controller-manager) |
2024-04-24 | v1.28.3-tke.3 | Fixed the Load Balancer vip binding issue. (kube-proxy) |
2024-04-10 | v1.28.3-tke.2 | Merged PR119561 to fix the issue where kubelet might panic when using gpu. (kubelet) Fixed the issue where Pods with eni-ip could not be scheduled to super nodes. (kube-scheduler) |
2024-03-06 | v1.28.3-tke.1 | Added support for specifying Pods during scale-in. (kube-controller-manager) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Added annotation support for mounting cgroupfs to Pods. (kubelet). Supported the native node scaling feature. (kube-scheduler) Correctly displayed disk and network metrics when the runtime is containerd. (kubelet) Super nodes support DaemonSets. (kube-apiserver/kube-controller-manager) Optimized HPA for super nodes. (kube-controller-manager). Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Changed the validity period of the apiserver's self-signed certificate from 1 year to 5 years. (kube-apiserver) Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). |
Date | Version | Update |
2026-07-22 | v1.26.1-tke.30 | Replace the kube-proxy base image to support iptables backend self-adaptation and ipset version compatibility, resolving abnormal issues caused by version mismatches. (Dockerfile) |
2026-06-29 | v1.26.1-tke.28 | Fixed the memory resource comparison logic by normalizing to bytes to ensure consistent comparison across different representations. (kubelet) kube-controller-manager supports progressive creation of DaemonSets. |
2026-03-24 | v1.26.1-tke.27 | Performed NodeAffinity filtering in the fixed IP (Eklet Static IP) pre-selection path and no longer ignored node affinity. (kube-scheduler) |
2026-01-27 | v1.26.1-tke.26 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) Added the EnableTKEXFeatures gate to support TKEx features. (kube-apiserver, kubelet, kube-scheduler) |
2026-01-06 | v1.26.1-tke.25 | Changed the default value of the MatchLabelKeysInPodTopologySpread feature gate to true. (kube-scheduler) Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-24 | v1.26.1-tke.24 | Fixed the abnormal issue with resource version tracking in the endpoints controller. (kube-controller-manager) |
2025-12-17 | v1.26.1-tke.23 | Fixed the issue where Pods encounter an UnexpectedAdmissionError when nodes in vpc-cni mode restart. (kubelet) During eviction, kube-controller-manager forcibly deletes pods on the failed node that have the annotation force-delete-when-node-unreachable=true. |
2025-09-08 | v1.26.1-tke.22 | Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) Supported disabling cadvisor reading proc fd through feature-gate. (kubelet) |
2025-08-07 | v1.26.1-tke.21 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.26.1-tke.20 | Optimized in-place kubelet configuration changes. (kubelet) Changed the validity period of cluster certificates generated by kubeadm from 1 year to 100 years. (kubeadm) Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-05-15 | v1.26.1-tke.19 | Optimized the reverse probing node of kcm. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.26.1-tke.18 | PR125675: Fixed the update inconsistency issue with endpoints. (kube-controller-manager) Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) Fixed the cache issue of monthly subscription super nodes in the scheduler. (kube-scheduler) Optimized the reverse probing node of kcm to parallel probing. (kube-controller-manager) Supported independent apiserver. (kube-apiserver) |
2025-04-01 | v1.26.1-tke.17 | Fixed the issue of illegal regular expression matching. (kube-apiserver) |
2025-03-25 | v1.26.1-tke.16 | PR3672: kubelet fixed the cadvisor cri client timeout issue. (kubelet) Supported precise speed limit for high-overhead list requests, one-click circuit-breaking systems, and business components such as List. (kube-apiserver) Supported configuring etcd overload protection via configmap. (kube-apiserver) PR121103: Fixed the GC issue. (kube-controller-manager) |
2025-03-06 | v1.26.1-tke.15 | Fixed the memory loss issue and data forwarding plane issue of kube-proxy. (kube-proxy) |
2025-02-25 | v1.26.1-tke.14 | Optimized kube-proxy performance. (kube-proxy) |
2025-01-09 | v1.26.1-tke.13 | Optimized resource awareness policies. (kube-scheduler) Mitigated apiserver cache breakdown. (kube-apiserver) |
2024-12-20 | v1.26.1-tke.12 | Fixed the issue where terminationGracePeriod did not meet expectations when kubelet deleted or drained Pods. (kubelet) Supported scale-out in a specified sequence by group. (kube-scheduler) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-11-27 | v1.26.1-tke.11 | Fixed the issue where static PV Pods could not be scheduled when the cluster had chc nodes. (kube-scheduler) |
2024-11-20 | v1.26.1-tke.10 | Fixed the cache issue with yearly/monthly subscription super nodes in the scheduler. (kube-scheduler) |
2024-09-24 | v1.26.1-tke.9 | Optimized the startup of the kube-proxy ipvs mode. (kube-proxy) |
2024-09-20 | v1.26.1-tke.8 | Optimized scheduler performance. (kube-scheduler) Supported in-place Pod specification upgrades. (kube-apiserver/kube-scheduler/kubelet) |
2024-07-09 | v1.26.1-tke.7 | Implemented etcd overload protection. (kube-apiserver) Optimized scheduling logic. (kube-scheduler) kube-proxy supported the sh+port scheduling method in ipvs mode. (kube-proxy) Enabled kubelet to support Pod deletion protection. (kubelet) |
2024-04-10 | v1.26.1-tke.6 | Merged PR119561 to fix the issue where kubelet might panic when using gpu. (kubelet) Merged PR113145 to fix the issue where a sandbox might be left behind during forced pod deletion, preventing pod recreation. (kubelet) |
2024-02-27 | v1.26.1-tke.5 | Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). |
2024-02-01 | v1.26.1-tke.4 | Super nodes support awareness of reserved specifications. (kube-scheduler) Merged PR119605 to fix the issue of incomplete pod information in the kubelet prober_probe_total metric. (kubelet) Optimized kube-proxy. (kube-proxy) job-controller supports configuring the multi-worker parameter --concurrent-job-syncs. (kube-controller-manager) |
2023-12-04 | v1.26.1-tke.3 | Added annotation support for mounting cgroupfs to Pods. (kubelet). Merged PR119706 to fix the issue where Pods using device plugin resources would be recreated upon kubelet restart. (kube-apiserver) Added userAgent reporting to metrics and detailed logging when resources are deleted. (kube-apiserver) |
2023-10-12 | v1.26.1-tke.2 | Supported in-place updates for Pod resources. (kube-apiserver,kubelet). Merged PR117550 to add the --concurrent-cron-job-syncs parameter for configuring the number of workers in the cronjob controller. (kube-controller-manager) Fixed the scheduler iterator error to prevent scores from exceeding the range of a single plugin due to duplicate scorePlugin additions. (kube-scheduler) Implemented overload protection for the apiserver. (kube-apiserver) Changed the validity period of self-signed certificates from 1 year to 5 years. (kube-apiserver) |
2023-03-30 | v1.26.1-tke.1 | Added support for specifying Pods during scale-in. (kube-controller-manager) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Fixed the issue with Load Balancer health checks and IPVS. (kube-proxy) Optimized HPA for super nodes. (kube-controller-manager). Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Added support for binding Load Balancer (LB) addresses to the ipvs network interface. (kube-proxy) Supported the native node scaling feature. (kube-scheduler) Enhanced the priority and fairness features of kube-apiserver. (kube-apiserver) Enhanced APF: sorted the output of dump_priority_levels by priority level (pl) name and added PL debug information. These changes have been merged into the community via PR112393. Added metrics for list/watch operations. (kube-apiserver) Added support for the exec/logs commands on super nodes and DaemonSet Pods. (kube-apiserver) Added metrics for network/disk and other resources for the containerd runtime. (kubelet) During pod garbage collection (gc), kube-controller-manager prioritizes the deletion of Evicted pods. (kube-controller-manager) |
Date | Version | Update |
2026-06-29 | v1.24.4-tke.39 | Fixed the memory resource comparison logic by normalizing to bytes to ensure consistent comparison across different representations. (kubelet) kube-controller-manager supports progressive creation of DaemonSets. |
2026-03-24 | v1.24.4-tke.38 | Performed NodeAffinity filtering in the fixed IP (Eklet Static IP) pre-selection path and no longer ignored node affinity. (kube-scheduler) |
2026-01-27 | v1.24.4-tke.37 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) |
2026-01-06 | v1.24.4-tke.36 | Fixed the issue where the scheduler's topology feature did not behave as expected. (kube-scheduler) Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-24 | v1.24.4-tke.35 | Fixed the abnormal issue with resource version tracking in the endpoints controller. (kube-controller-manager) |
2025-12-17 | v1.24.4-tke.34 | During eviction, kube-controller-manager forcibly deletes pods on the failed node that have the annotation force-delete-when-node-unreachable=true. |
2025-11-19 | v1.24.4-tke.33 | Fixed the EndpointSlice hash to ensure that different Pods with the same IP address were exposed by Service. (kube-controller-manager) Fixed the issue where concurrent write operations to a map in the scheduler caused panic. (kube-scheduler) |
2025-09-08 | v1.24.4-tke.32 | Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) Supported disabling cadvisor reading proc fd through feature-gate. (kubelet) |
2025-08-07 | v1.24.4-tke.31 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.24.4-tke.30 | Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-07-09 | v1.24.4-tke.29 | Optimized in-place kubelet configuration changes. (kubelet) Changed the validity period of cluster certificates generated by kubeadm from 1 year to 100 years. (kubeadm) |
2025-05-15 | v1.24.4-tke.28 | Optimized the reverse probing node of kcm. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.24.4-tke.27 | PR125675: Fixed the update inconsistency issue with endpoints. (kube-controller-manager) Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) Fixed the cache issue of monthly subscription super nodes in the scheduler. (kube-scheduler) Optimized the reverse probing node of kcm to parallel probing. (kube-controller-manager) Supported independent apiserver. (kube-apiserver) PR114168: A node was not forced to detach its mounted CSI volume when it was healthy. (kube-controller-manager) |
2025-04-01 | v1.24.4-tke.26 | Fixed the issue of illegal regular expression matching. (kube-apiserver) |
2025-03-25 | v1.24.4-tke.25 | PR3672: kubelet fixed the cadvisor cri client timeout issue. (kubelet) Supported precise speed limit for high-overhead list requests, one-click circuit-breaking systems, and business components such as List. (kube-apiserver) Supported configuring etcd overload protection via configmap. (kube-apiserver) |
2025-03-06 | v1.24.4-tke.24 | Fixed the memory loss issue and data forwarding plane issue of kube-proxy. (kube-proxy) |
2025-02-25 | v1.24.4-tke.23 | Optimized kube-proxy performance. (kube-proxy) |
2025-02-20 | v1.24.4-tke.22 | PR112607: Fixed the issue where the automatic binding of the kubelet directory to the data disk resulted in residual CBS mount points. (kubelet) |
2025-01-09 | v1.24.4-tke.21 | Allowed Pods to add or delete CPU or memory resources while ensuring Pod QOSClass remained unchanged. Allowed Pods to reduce limits. (apiserver) Optimized resource awareness policies. (kube-scheduler) Mitigated apiserver cache breakdown. (kube-apiserver) |
2024-12-20 | v1.24.4-tke.20 | Fixed the issue where terminationGracePeriod did not meet expectations when kubelet deleted or drained Pods. (kubelet) Optimized in-place specification upgrades. (apiserver/scheduler/kubelet) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-11-27 | v1.24.4-tke.19 | Fixed the issue where static PV Pods could not be scheduled when the cluster had chc nodes. (kube-scheduler) |
2024-11-20 | v1.24.4-tke.18 | Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) Enabled the scheduler to scale out in a specified sequence by group. (kube-scheduler) Fixed the crash issue with HPA version conversion. (kube-controller-manager) Supported setting the number of local replicas to 0 for yearly/monthly subscription nodes. (kube-scheduler) Fixed the cache issue with yearly/monthly subscription super nodes in the scheduler. (kube-scheduler) |
2024-09-24 | v1.24.4-tke.17 | Optimized the startup of the kube-proxy ipvs mode. (kube-proxy) |
2024-07-02 | v1.24.4-tke.16 | Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) |
2024-07-09 | v1.24.4-tke.15 | Implemented etcd overload protection. (kube-apiserver) Optimized scheduling logic. (kube-scheduler) kube-proxy supported the sh+port scheduling method in ipvs mode. (kube-proxy) Enabled kubelet to support Pod deletion protection. (kubelet) |
2024-04-10 | v1.24.4-tke.14 | Merged PR119561 to fix the issue where kubelet might panic when using gpu. (kubelet) Merged PR113145 to fix the issue where a sandbox might be left behind during forced pod deletion, preventing pod recreation. (kubelet) |
2024-02-27 | v1.24.4-tke.13 | Optimized sandbox retention for super nodes. (kube-scheduler) Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). Optimized the compute resource policy for super nodes. (kube-scheduler) |
2024-02-01 | v1.24.4-tke.12 | Optimized kube-proxy. (kube-proxy) job-controller supports configuring the multi-worker parameter --concurrent-job-syncs. (kube-controller-manager) |
2023-12-04 | v1.24.4-tke.11 | Added annotation support for mounting cgroupfs to Pods. (kubelet). Added userAgent reporting to metrics and detailed logging when resources are deleted. (kube-apiserver) |
2023-10-12 | v1.24.4-tke.10 | Backported CVE-2022-3162. (kube-apiserver) Implemented overload protection for the apiserver. (kube-apiserver) The validity period of self-signed certificates has been changed from 1 year to 5 years. (kube-apiserver) Merged PR116327 to optimize memory usage for GetList. (kube-apiserver) Added monitoring metric collection. (kube-scheduler) |
2023-07-21 | v1.24.4-tke.9 | Enabled kube-proxy to detect changes in node PodCIDRs and restart to ensure consistency when changes occur. (kube-proxy) Fixed the issue where kubelet might panic when assigning devices to Pods. (kubelet) Fixed the scheduler iterator error to prevent scores from exceeding the range of a single plugin due to duplicate scorePlugin additions. (kube-scheduler) Backported CVE-2022-3294. (kube-apiserver) |
2023-05-30 | v1.24.4-tke.7 | Supported in-place updates for Pod resources. (kube-apiserver,kube-scheduler,kubelet) Optimized scheduling logs. (kube-scheduler) Merged PR117550 to add the --concurrent-cron-job-syncs parameter for configuring the number of workers in the cronjob controller. (kube-controller-manager) |
2023-03-17 | v1.24.4-tke.6 | Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). Added support for specifying Pods during scale-in. (kube-controller-manager) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Merged PR110951 to prevent the issue where a PV mount operation occasionally hangs, causing all Pods associated with that PV to be stuck in the Creating state. (kubelet) Merged PR111635 to fix the issue where ipvs real server weights become zero after kube-proxy restarts. (kube-proxy) Supported the native node scaling feature. (kube-scheduler) Added support for binding Load Balancer (LB) addresses to the ipvs network interface. (kube-proxy) Enhanced priority and fairness features for kube-apiserver: (kube-apiserver)
a. Added support for rate limiting based on userAgent, token bucket rate limiting, and rate limiting and circuit breaking for any type of component.
b. Sorted by pl name when priority levels are dumped; added PL debug information: total processed requests, total rejected requests, total timed-out requests, total canceled requests. Record metrics when list/watch operations encounter too old, too large, or too many objects (kube-apiserver): watch_too_old_objects_events_total, list_too_large_objects_events_total, list_too_many_objects_events_total, watch_too_many_objects_events_totalAdded support for network metrics in the containerd runtime. (kubelet) Added support for Tencent Cloud CBS CSI migration. (kube-apiserver, kube-controller-manager, kubelet) During pod garbage collection (gc), kube-controller-manager prioritizes the deletion of Evicted pods. (kube-controller-manager) |
2023-02-16 | v1.24.4-tke.5 | Added support for the exec/logs commands on super nodes and DaemonSet Pods. (kube-apiserver). Added support for fixed IPs on super nodes. (kube-scheduler) Backported the support for multiple HPAController workers from version 1.26. (kube-controller-manager) Backported the community's heuristic feature for setting watch channel sizes and added support for configuring watch channel sizes for specific resources via environment variables. (kube-apiserver). Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) |
2022-12-08 | v1.24.4-tke.3 | Added support for disk usage metrics in the containerd runtime. (kubelet) Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Added support for TKE Serverless super nodes. (kube-controller-manager, kube-scheduler) |
2022-05-07 | v1.24.4-tke.1 | Updated the startup method for kube-proxy when kube-proxy is running in image mode to automatically adapt to the iptables mode of the node, thereby supporting operating systems that use nf_tables mode by default for iptables. (kube-proxy) Allowed the use of special CIDR blocks for TKE managed clusters. (kube-apiserver) Reverted PR63066 to fix the issue with Load Balancer health checks and IPVS. (kube-proxy) Merged PR112299 to optimize the gzip compression level for the apiserver. |
Date | Version | Update |
2026-06-29 | v1.22.5-tke.48 | Fixed the memory resource comparison logic by normalizing to bytes to ensure consistent comparison across different representations. (kubelet) kube-controller-manager supports progressive creation of DaemonSets. |
2026-03-24 | v1.22.5-tke.47 | Performed NodeAffinity filtering in the fixed IP (Eklet Static IP) pre-selection path and no longer ignored node affinity. (kube-scheduler) |
2026-02-26 | v1.22.5-tke.46 | Fixed the issue where the kube-scheduler topology feature did not behave as expected. (kube-scheduler). |
2026-01-27 | v1.22.5-tke.45 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) |
2026-01-06 | v1.22.5-tke.44 | Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-24 | v1.22.5-tke.43 | Fixed the abnormal issue with resource version tracking in the endpoints controller. (kube-controller-manager) |
2025-12-17 | v1.22.5-tke.42 | During eviction, kube-controller-manager forcibly deletes pods on the failed node that have the annotation force-delete-when-node-unreachable=true. |
2025-11-19 | v1.22.5-tke.41 | Fixed the EndpointSlice hash to ensure that different Pods with the same IP address were exposed by Service. (kube-controller-manager) Fixed the issue where concurrent write operations to a map in the scheduler caused panic. (kube-scheduler) |
2025-09-08 | v1.22.5-tke.40 | Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) Supported disabling cadvisor reading proc fd through feature-gate. (kubelet) |
2025-08-07 | v1.22.5-tke.39 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.22.5-tke.38 | Optimized in-place kubelet configuration changes. (kubelet) Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-05-15 | v1.22.5-tke.37 | Optimized the reverse probing node of kcm. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.22.5-tke.36 | PR125675: Fixed the update inconsistency issue with endpoints. (kube-controller-manager) Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) Fixed the cache issue of monthly subscription super nodes in the scheduler. (kube-scheduler) Optimized the reverse probing node of kcm to parallel probing. (kube-controller-manager) Supported independent apiserver. (kube-apiserver) |
2025-04-01 | v1.22.5-tke.35 | Fixed the issue of illegal regular expression matching. (kube-apiserver) |
2025-03-25 | v1.22.5-tke.34 | PR3672: kubelet fixed the cadvisor cri client timeout issue. (kubelet) Supported precise speed limit for high-overhead list requests, one-click circuit-breaking systems, and business components such as List. (kube-apiserver) Supported configuring etcd overload protection via configmap. (kube-apiserver) |
2025-03-06 | v1.22.5-tke.33 | Fixed the memory loss issue and data forwarding plane issue of kube-proxy. (kube-proxy) |
2025-02-25 | v1.22.5-tke.32 | PR108325: Fixed the issue where kubelet did not report an error but only printed logs when the sandbox container did not exist during container manager startup. (kubelet) |
2025-01-09 | v1.22.5-tke.31 | Allowed Pods to add or delete CPU or memory resources while ensuring Pod QOSClass remained unchanged. Allowed Pods to reduce limits. (apiserver) Optimized resource awareness policies. (kube-scheduler) Mitigated apiserver cache breakdown. (kube-apiserver) |
2024-12-20 | v1.22.5-tke.30 | Fixed the issue where terminationGracePeriod did not meet expectations when kubelet deleted or drained Pods. (kubelet) Optimized in-place specification upgrades. (apiserver/scheduler/kubelet) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-11-27 | v1.22.5-tke.29 | Fixed the issue where static PV Pods could not be scheduled when the cluster had chc nodes. (kube-scheduler) |
2024-11-20 | v1.22.5-tke.28 | Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) Enhanced endpointslice controller metrics cache performance. (kube-controller-manager) Optimized the reverse probing node of kcm. (kube-controller-manager) |
2024-09-24 | v1.22.5-tke.27 | Optimized the startup of the kube-proxy ipvs mode. (kube-proxy) |
2024-09-24 | v1.22.5-tke.26 | Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) |
2024-07-09 | v1.22.5-tke.25 | Implemented etcd overload protection. (kube-apiserver) Optimized scheduling logic. (kube-scheduler) Enabled kubelet to support Pod deletion protection. (kubelet) |
2024-04-10 | v1.22.5-tke.24 | Merged PR119561 to fix the issue where kubelet might panic when using gpu. (kubelet) Merged PR106570 to fix the issue of setting cpushare in kubelet. (kubelet) |
2024-02-27 | v1.22.5-tke.23 | Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). Optimized the compute resource policy for super nodes. (kube-scheduler) |
2024-02-01 | v1.22.5-tke.22 | job-controller supports configuring the multi-worker parameter --concurrent-job-syncs. (kube-controller-manager) |
2023-12-04 | v1.22.5-tke.21 | Backported PR108753 to fix the issue where pod memory and other metrics could not be obtained via kubelet /stats/summary when systemd cgroup is used. (kubelet) Added annotation support for mounting cgroupfs to Pods. (kubelet). Added userAgent reporting to metrics and detailed logging when resources are deleted. (kube-apiserver) |
2023-10-12 | v1.22.5-tke.20 | Changed the validity period of self-signed certificates from 1 year to 5 years. (kube-apiserver) Merged PR116327 to optimize memory usage for GetList. (kube-apiserver) Merged PR107900 to fix the issue where static pods could not be started in certain scenarios. (kubelet) Reverted the fix for the issue caused by the ubuntu16.04lxcfs upgrade, as the fix is no longer needed and would cause e2e test failures. (kubelet) Backported CVE-2022-3162. (kube-apiserver) |
2023-08-02 | v1.22.5-tke.19 | Implemented overload protection for the apiserver. (kube-apiserver) |
2023-07-21 | v1.22.5-tke.18 | Fixed the issue where kubelet might panic when assigning devices to Pods. (kubelet) Fixed the bug in the super node custom policy. (kube-scheduler) Backported CVE-2022-3294. (kube-apiserver) |
2023-05-26 | v1.22.5-tke.16 | Backported PR105606&PR105872 to fix the goroutine leak issue caused by DeleteCollection. (kube-apiserver) Backported PR117550 to add the --concurrent-cron-job-syncs parameter for configuring the number of workers in the cronjob controller. (kube-controller-manager) |
2023-04-20 | v1.22.5-tke.15 | When cadvisor collects metrics, it skips the process if the files for a newly created container are not yet available, preventing gaps in reported data caused by empty values. (kubelet) Merged PR112913 to fix the issue where the cpu manager calculates the allocatable cpuset. (kubelet) Improved the handling logic for in-place Pod specification changes. (kube-apiserver,kube-scheduler,kubelet) Merged PR107670 to fix the issue where a timing Bug might be triggered during Pod deletion, leading to sandbox leakage. (kubelet) |
2023-03-24 | v1.22.5-tke.14 | Merged PR108202 to fix the issue where the nodeName field in EndpointSlice was missing after an upgrade from version 1.20 to 1.22 is performed. (kube-apiserver) Merged PR110245 to add a fallback logic in kube-proxy that uses endpoint.DeprecatedTopology[v1.LabelHostname] when endpoint.NodeName is empty, preventing potential service inaccessibility after an upgrade from version 1.20 to 1.22. (kube-proxy) |
2023-03-17 | v1.22.5-tke.13 | Merged PR110951 to prevent the issue where a PV mount operation occasionally hangs, causing all Pods associated with that PV to be stuck in the Creating state. (kubelet) Fixed the issue where GET requests were not logged as WATCH operations under certain circumstances. (kube-apiserver) During pod garbage collection (gc), kube-controller-manager prioritizes the deletion of Evicted pods. (kube-controller-manager) Added support for binding Load Balancer (LB) addresses to the ipvs network interface. (kube-proxy) |
2023-03-06 | v1.22.5-tke.12 | Merged PR111635 to fix the issue where ipvs real server weights become zero after kube-proxy restarts. (kube-proxy) |
2023-02-16 | v1.22.5-tke.11 | Added support for fixed IPs on super nodes. (kube-scheduler) Backported the support for multiple HPAController workers from version 1.26. (kube-controller-manager) Backported the community's heuristic feature for setting watch channel sizes and added support for configuring watch channel sizes for specific resources via environment variables. (kube-apiserver). Backported PR108366 to fix the issue where kubelet prematurely reports the status of terminating pods, causing new pods to encounter insufficient resources during runtime. (kubelet) Fixed the issue where inotify leaks cause kubelet memory leaks. (kubelet) Removed logs related to watch_too_old_objects_events_total, list_too_many_objects_events_total, and watch_too_many_objects_events_total to avoid excessive information output. (kube-apiserver) Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) |
2023-01-12 | v1.22.5-tke.9 | Supported the native node scaling feature. (kube-scheduler) Added a delegated_to_storage Tag to the list_too_many_objects_events_total metric to indicate whether the request is delegated to etcd. (kube-apiserver) |
2023-01-05 | v1.22.5-tke.8 | Enhanced the priority and fairness features of kube-apiserver. (kube-apiserver) a. During rate limiting, the number of objects of the resource type in list requests is considered. A higher number of objects results in lower concurrent QPS, significantly reducing the resources occupied by the APIServer, effectively enhancing cluster stability and preventing cascading failures. If business components issue a large number of list requests, latency may increase after the upgrade, and the APIServer may return rate-limiting retries. b. Supports rate limiting based on userAgent, token bucket rate limiting, and rate limiting and circuit breaking for any type of component. c. Supports rate limiting for update requests related to objects in a watch, preventing a sudden surge of update events. Super nodes support running DaemonSet Pods when needed. (kube-controller-manager). Added support for the exec/logs commands on super nodes and DaemonSet Pods. (kube-apiserver). Records metrics when list/watch operations encounter too old, too large, or too many objects (kube-apiserver): watch_too_old_objects_events_total, list_too_large_objects_events_total, list_too_many_objects_events_total, watch_too_many_objects_events_totalMerged PR108754 to fix the issue where an unregistered error occasionally occurs when ConfigMap/Secrets volumes are mounted. (kubelet) Supported in-place updates for Pod resources. (kube-apiserver,kubelet). |
2022-12-08 | v1.22.5-tke.7 | Fixed the issue where the default name of the container network interface was set to eth1 (kubelet). |
2022-11-09 | v1.22.5-tke.6 | Optimized the scheduling of yearly/monthly subscription super nodes. (kube-scheduler) Optimized the stability of cloud native nodes. (kubelet) Optimized the scheduling of local replica counts for super nodes. Added the QosIgnoreInitContainer option. When enabled, it ignores init container resources when pod QoS is set. (kube-apiserver,kubelet) |
2022-10-13 | v1.22.5-tke.5 | Serverless clusters support DaemonSets. (kube-apiserver). Merged PR112299 to optimize the gzip compression level for the apiserver. (kube-apiserver). Fixed CVE-2022-3172. (kube-apiserver) Added the calling_webhook_timeout_error metric. (kube-apiserver). Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). Merged PR110294 to fix the issue where the Job activeDeadlineSeconds setting does not take effect. (kube-controller-manager) Merged PR111773 to fix the memory leak issue during scheduler preemption. (kube-scheduler) Added support for containerd disk metrics. (kubelet) Merged PR108831 to fix the kubelet panic issue caused by creating multiple Pods simultaneously. (kubelet) Added support for specifying Pods during scale-in. (kube-controller-manager) Optimized daemonset pod scheduling performance by processing only the nodes assigned to them. (kube-scheduler) Set the CBS CSI Migration to the completed state. (kube-controller-manager) Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Optimized HPA for super nodes in Serverless clusters. (kube-controller-manager) |
2022-05-07 | v1.22.5-tke.1 | Allowed the use of special CIDR blocks for TKE managed clusters. (kube-apiserver) Reverted PR63066 to fix the issue with Load Balancer health checks and IPVS. (kube-proxy) Merged PR90260 to fix the missing network monitoring issue for containerd clusters. (kubelet) Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. (kubelet) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Added support for Tencent Cloud CBS CSI Migration. (kube-controller-manager, kubelet) Merged PR106906 to detect whether the subpath of the network storage volume has been deleted, preventing Pods from remaining in the terminating state. (kubelet) Updated the startup method for kube-proxy when kube-proxy is running in image mode to automatically adapt to the iptables mode of the node, thereby supporting operating systems that use nf_tables mode by default for iptables. (kube-proxy) |
Date | Version | Update |
2026-06-29 | v1.20.6-tke.66 | Fixed the memory resource comparison logic by normalizing to bytes to ensure consistent comparison across different representations. (kubelet) kube-controller-manager supports progressive creation of DaemonSets. |
2026-03-24 | v1.20.6-tke.65 | Performed NodeAffinity filtering in the fixed IP (Eklet Static IP) pre-selection path and no longer ignored node affinity. (kube-scheduler) |
2026-02-26 | v1.20.6-tke.64 | Fixed the issue where the kube-scheduler topology feature did not behave as expected. (kube-scheduler). |
2026-01-27 | v1.20.6-tke.63 | Fixed the issue of arm64 architecture nodes failing to join the cluster. (kubelet) Added the DisableFdSocketMetrics feature gate (disabled by default). When enabled, it prevents the collection of fd and socket information. (kubelet). |
2026-01-06 | v1.20.6-tke.62 | Added an eklet node check to the kcm process pod logic. (kube-controller-manager) |
2025-12-24 | v1.20.6-tke.61 | Fixed the abnormal issue with resource version tracking in the endpoints controller. (kube-controller-manager) |
2025-11-19 | v1.20.6-tke.60 | Fixed the EndpointSlice hash to ensure that different Pods with the same IP address were exposed by Service. (kube-controller-manager) |
2025-09-08 | v1.20.6-tke.59 | Optimized the self-adaptive speed limit logic of apiserver. (kube-apiserver) |
2025-08-07 | v1.20.6-tke.58 | Enabled apiserver to support self-adaptive traffic control. (kube-apiserver) |
2025-07-18 | v1.20.6-tke.57 | Optimized in-place kubelet configuration changes. (kubelet) Enabled apiserver to support the stream-list feature. (kube-apiserver) |
2025-05-15 | v1.20.6-tke.56 | Optimized the reverse probing node of kcm. For details about this feature, refer to TKE Kubernetes Feature Description. (kube-controller-manager) |
2025-05-12 | v1.20.6-tke.55 | PR125675: Fixed the update inconsistency issue with endpoints. (kube-controller-manager) Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) Fixed the cache issue of monthly subscription super nodes in the scheduler. (kube-scheduler) Optimized the reverse probing node of kcm to parallel probing. (kube-controller-manager) Supported independent apiserver. (kube-apiserver) |
2025-04-01 | v1.20.6-tke.54 | Fixed the issue of illegal regular expression matching. (kube-apiserver) |
2025-03-25 | v1.20.6-tke.53 | PR99839: Fixed the leak issue caused by kubelet port forwarding. (kubelet) Supported precise speed limit for high-overhead list requests, one-click circuit-breaking systems, and business components such as List. (kube-apiserver) Supported configuring etcd overload protection via configmap. (kube-apiserver) |
2025-03-06 | v1.20.6-tke.52 | Fixed the memory loss issue and data forwarding plane issue of kube-proxy. (kube-proxy) |
2025-02-25 | v1.20.6-tke.51 | Optimized kube-proxy performance. (kube-proxy) |
2025-01-09 | v1.20.6-tke.50 | PR106570: Fixed the bug of setting cpushare in kubelet. (kubelet) Allowed Pods to add or delete CPU or memory resources while ensuring Pod QOSClass remained unchanged. Allowed Pods to reduce limits. (apiserver) Optimized resource awareness policies. (kube-scheduler) Mitigated apiserver cache breakdown. (kube-apiserver) |
2024-12-20 | v1.20.6-tke.49 | Optimized in-place specification upgrades. (apiserver/scheduler/kubelet) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-11-27 | v1.20.6-tke.48 | Fixed the issue where static PV Pods could not be scheduled when the cluster had chc nodes. (kube-scheduler) |
2024-11-20 | v1.20.6-tke.47 | Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) kube-proxy supported removing tcp conntrack endpoints. (kube-proxy) Fixed the issue where expired Pods occupied resources. (kube-scheduler) Enhanced endpointslice controller metrics cache performance. (kube-controller-manager) PR96582: Fixed the duplicate scheduling issue in certain scenarios. (kube-scheduler) Enabled the scheduler to scale out in a specified sequence by group. (kube-scheduler) Fixed the crash issue with HPA version conversion. (kube-controller-manager) Supported setting the number of local replicas to 0 for monthly subscription native nodes. (kube-scheduler) Fixed the issue where the terminationGracePeriod did not meet expectations when kubelet deleted or drained Pods. (kubelet) |
2024-09-24 | v1.20.6-tke.46 | Optimized the startup of the kube-proxy ipvs mode. (kube-proxy) |
2024-09-20 | v1.20.6-tke.45 | Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) |
2024-07-09 | v1.20.6-tke.44 | Implemented etcd overload protection. (kube-apiserver) Optimized scheduling logic. (kube-scheduler) Enabled kubelet to support Pod deletion protection. (kubelet) |
2024-04-10 | v1.20.6-tke.43 | Allows customizing the eviction delay to determine the AZ eviction sequence. (kube-scheduler) |
2024-02-27 | v1.20.6-tke.42 | Optimized sandbox retention for super nodes. (kube-scheduler) Optimized the performance of workload-mounted disks. (kube-controller-manager) Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). Optimized the compute resource policy for super nodes. (kube-scheduler) |
2024-02-01 | v1.20.6-tke.41 | Native nodes support CVM mode. (kubelet). Merged PR102014 to fix the issue where CPUs bound to init containers became unusable after release. (kubelet). job-controller supports configuring the multi-worker parameter --concurrent-job-syncs. (kube-controller-manager) |
2023-12-04 | v1.20.6-tke.40 | Added annotation support for mounting cgroupfs to Pods. (kubelet). Optimized scheduler patch pods. (kube-scheduler) Added userAgent reporting to metrics and detailed logging when resources are deleted. (kube-apiserver) |
2023-10-12 | v1.20.6-tke.39 | Changed the validity period of self-signed certificates from 1 year to 5 years. (kube-apiserver) Reverted the fix for the issue caused by the ubuntu16.04lxcfs upgrade, as the fix is no longer needed and would cause e2e test failures. (kubelet) |
2023-08-02 | v1.20.6-tke.38 | Merged PR103320 and bumped SMD to v4.1.2 to pick up the #102749 fix. Implemented overload protection for the apiserver. (kube-apiserver) |
2023-07-20 | v1.20.6-tke.37 | Merged PR116327 to optimize memory usage for List. (kube-apiserver) Made the qps/burst of the hpa controller configurable. (kube-controller-manager) Backported PR107831 to include pod UID in the kubelet secret/configmap cache. (kubelet). Hardened cadvisor collection. (kubelet) Optimized the performance of the attach detach controller to accelerate processing speed during high-frequency node updates. (kube-controller-manager) Merged CVE-2022-3294. (kube-apiserver) Fixed the bug in the super node custom policy. (kube-scheduler) |
2023-05-26 | v1.20.6-tke.34 | Fixed the issue where the watch_cache_capacity metric resource names for CRD instances were all "Unstructured", causing indistinguishability and mutual interference. (kube-apiserver) Backported PR116167. During pod garbage collection (gc), kube-controller-manager prioritizes the deletion of Evicted pods. (kube-controller-manager) Backported PR97845 and added the apiserver_terminated_watchers_total metric, which indicates the number of watchers closed due to untimely processing. (kube-apiserver) Backported PR105606&PR105872 to fix the goroutine leak issue caused by DeleteCollection. (kube-apiserver) Backported PR117550 to add the --concurrent-cron-job-syncs parameter for configuring the number of workers in the cronjob controller. (kube-controller-manager) Improved the handling logic for in-place Pod specification changes. (kube-apiserver,kube-scheduler,kubelet) |
2023-03-10 | v1.20.6-tke.33 | Merged PR110951 to prevent the issue where a PV mount operation occasionally hangs, causing all Pods associated with that PV to be stuck in the Creating state. (kubelet) Merged PR106295 to enable the Device Manager to support both v1 and v2 formats when the Device Manager reads state files. (kubelet) Fixed the issue where GET requests were not logged as WATCH operations under certain circumstances. (kube-apiserver) |
2023-03-06 | v1.20.6-tke.32 | Merged PR111635, fixing the issue where the ipvs real server weight becomes 0 after kube-proxy restarts. (kube-proxy) |
2023-02-13 | v1.20.6-tke.31 | Backported the support for multiple HPAController workers from version 1.26. (kube-controller-manager) Backported the community's heuristic feature for setting watch channel sizes and added support for configuring watch channel sizes for specific resources via environment variables. (kube-apiserver). Fixed the issue where inotify leaks cause kubelet memory leaks. (kubelet) Removed logs related to watch_too_old_objects_events_total, list_too_many_objects_events_total, and watch_too_many_objects_events_total to avoid excessive information output. (kube-apiserver) Score super nodes based on the eks.tke.cloud.tencent.com/eklet-priority annotation. (kube-scheduler)Fixed the bug related to static IP addresses for super nodes. (kube-scheduler) Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) Fixed the issue where the rest_client_rate_limiter_duration_seconds metric was not registered. (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, kube-proxy) |
2023-01-05 | v1.20.6-tke.30 | Merged PR104833 to optimize the controller lock in API Priority and Fairness to a read-write lock. (kube-apiserver) Super nodes support running DaemonSet Pods when needed. (kube-controller-manager). Merged PR102576 to stop using RemoveAll when leftover Pod volumes are deleted, preventing the deletion of volume-related data. (kubelet) Allows DaemonSet Pods with specific annotations to be scheduled onto super nodes. (kube-controller-manager) Added a switch to enable or disable the local replica quantity policy for super nodes. (kube-scheduler) Enhanced the priority and fairness features of kube-apiserver. (kube-apiserver) a. During rate limiting, the number of objects of the resource type in list requests is considered. A higher number of objects results in lower concurrent QPS, significantly reducing the resources occupied by the APIServer, effectively enhancing cluster stability and preventing cascading failures. If business components issue a large number of list requests, latency may increase after the upgrade, and the APIServer may return rate-limiting retries. b. Supports rate limiting based on userAgent, token bucket rate limiting, and rate limiting and circuit breaking for any type of component. c. Supports rate limiting for update requests related to objects in a watch, preventing a sudden surge of update events. Added support for fixed IPs on super nodes. (kube-scheduler) Added a switch to enable or disable features related to super nodes. (kube-scheduler) Records metrics when list/watch operations encounter too old, too large, or too many objects (kube-apiserver): watch_too_old_objects_events_total, list_too_large_objects_events_total, list_too_many_objects_events_total, watch_too_many_objects_events_totalAdded a capacity metric for the Watch cache: watch_cache_capacity. (kube-apiserver) Optimized super node scheduling performance using ConfigMap Informer. (kube-scheduler) Supports setting the minimum value for the watch cache quantity via the environment variable DEFAULT_LOWER_BOUND_CAPACITY. (kube-apiserver) Added the AllowLocalConnection switch to enable external requests to access local addresses when it is enabled. (kubelet) |
2022-11-09 | v1.20.6-tke.28 | Optimized the scheduling of yearly/monthly subscription super nodes. (kube-scheduler) Optimized the stability of cloud native nodes. (kubelet) Fixed the abnormal exit of kubelet during in-place specification adjustment of native nodes. (kubelet) Optimized the scheduling of local replica counts for super nodes. (kube-scheduler) Added the QosIgnoreInitContainer option. When enabled, it ignores init container resources when pod QoS is set. (kube-apiserver,kubelet) |
2022-10-13 | v1.20.6-tke.27 | Optimized the scheduling policy for super node IP address management. (kube-scheduler) Merged PR112299 to optimize the gzip compression level for the apiserver. (kube-apiserver). Fixed CVE-2022-3172. (kube-apiserver) Added the calling_webhook_timeout_error metric. (kube-apiserver). Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). Optimized the scheduling of local replica counts for Serverless clusters. (kube-scheduler) Removed the forced scheduling of fixed EIPs to their original super nodes. (kube-scheduler) Merged PR99324. The token is immediately reset after it expires for the kube-controller-manager. (kube-controller-manager) Merged PR101155 to support configuring multiple ServiceAccount token issuers. (kube-apiserver) Added the StatusDetaching state for CBS disks. (kube-controller-manager) Increased the timeout for resource updates during static IP address scheduling for super nodes to prevent update failures during large-scale concurrent operations. (kube-scheduler) Serverless clusters support DaemonSets. (kube-apiserver). |
2022-09-07 | v1.20.6-tke.24 | Optimized scheduler preemption to avoid crashes. (kube-scheduler) Optimized super node scheduling. (kube-scheduler). Supported in-place updates for Pod resources. (kube-apiserver,kubelet). Supported scheduling Pods in the kube-system namespace to prepaid super nodes. (kube-scheduler). Optimized HPA for super nodes. (kube-controller-manager). Merged PR110294 to fix the issue where the Job activeDeadlineSeconds setting does not take effect. (kube-controller-manager) Merged PR111773 to fix the memory leak issue during scheduler preemption. (kube-scheduler) Merged PR97348 to fix the issue where HPA scaling quantities were incorrect when StabilizationWindowSeconds was set. (kube-controller-manager). Optimized the scheduling of yearly/monthly subscription super nodes. Merged PR108831 to fix the kubelet panic issue caused by creating multiple Pods simultaneously. (kubelet). Fixed the issue where Pod creation failed when the Pod name/uid of a cronjob was empty. (kube-controller-manager) |
2022-07-27 | v1.20.6-tke.21 | CBS supports native nodes. (kubelet). Optimized HPA for TKE Serverless super nodes. |
2022-06-16 | v1.20.6-tke.20 | Supports yearly/monthly subscription super nodes. (kube-scheduler, kube-controller-manager) When docker and overlay2 are used, disk usage is obtained via fs quota to improve performance. (kubelet) Optimized daemonset pod scheduling performance by processing only the nodes assigned to them. (kube-scheduler) Optimized TKE Serverless scheduling. (kube-scheduler). TKE Serverless: Supports mounting PVCs after pod creation. (kube-scheduler) When pods are scheduled to TKE Serverless nodes, hugepages resources can be ignored via a feature flag. (kube-scheduler) |
2022-04-22 | v1.20.6-tke.17 | TKE Serverless: Supports the sandbox retention feature. (kube-scheduler) Merged PR101093 to prevent the issue where the startupProbe is not executed after a container restart. (kubelet) |
2022-03-24 | v1.20.6-tke.16 | Fixed the issue where inline csi and ephemeral generic temporary volumes become unavailable after an upgrade to version 1.20 is performed. (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, kubectl) |
2022-03-18 | v1.20.6-tke.15 | Supports specifying pods during scale-in. (kube-controller-manager) Merged PR106906 to detect whether the subpath of the network storage volume has been deleted, preventing Pods from remaining in the terminating state. (kubelet) When anti-affinity scheduling is performed based on hostname, TKE Serverless super nodes are ignored. (kube-scheduler) Supports upgrading from TKE 1.18 to 1.20. (kube-apiserver, kube-controller-manager, kubelet) Backported PR108325 to fix the issue where kubelet might panic during startup due to the pause container being deleted. (kubelet) Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). |
2022-01-20 | v1.20.6-tke.12 | TKE Serverless rescheduling optimization: Reduces the score for super nodes within the same AZ that have been evicted. (kube-scheduler) The apiserver supports integrating external services of the ExternalName type. (kube-apiserver) Added support for binding Load Balancer (LB) addresses to the ipvs network interface. (kube-proxy) |
2021-12-09 | v1.20.6-tke.9 | Optimized scheduling and HPA for TKE Serverless super nodes. (kube-controller-manager, kube-scheduler) Fixed the issue where TKE Serverless cpu resource calculations were inconsistent with the frontend. (kube-scheduler) |
2021-12-02 | v1.20.6-tke.8 | Optimized grpc logs to prevent excessive logging when kubelet collects volume status. (kubelet) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) |
2021-11-26 | v1.20.6-tke.7 | Added support for custom installation of other CNIs on hybrid cloud external nodes. (kube-controller-manager) Avoided unnecessary processing of updates after a Pod is Assumed. (kube-scheduler) Merged PR99336 to improve the synchronization mechanism for node information during kubelet startup. (kubelet) |
2021-10-13 | v1.20.6-tke.6 | Merged PR89465, fixing the issue where HPA incorrectly calculates the number of instances based on pod metrics during rolling updates. (kube-controller-manager) |
2021-09-27 | v1.20.6-tke.5 | Supports collecting disk usage metrics for the containerd runtime. (kubelet) |
2021-09-23 | v1.20.6-tke.4 | Fixed the issue where storage metrics had no data when cgroup v2 is used. (kubelet) Fixed CVE-2021-25741 to prevent unauthorized access to host files via symbolic links. (kubelet) |
2021-07-19 | v1.20.6-tke.3 | When scaling out from a TKE cluster to TKE Serverless is performed, the scheduler correctly perceives the remaining IPs in the subnet during batch Pod scheduling, and schedules the appropriate number of Pods to super nodes. (kube-scheduler) Port upstream modifications for kubelet and cadvisor to fix issues with metric collection and statistics when cgroupv2 is used. (kubelet) |
2021-06-21 | v1.20.6-tke.2 | CSIMigration and CSIMigrationQcloudCbs are enabled by default, mounting CBS disks in the CSI manner. |
2021-05-25 | v1.20.6-tke.1 | Reverted PR63066 to fix the issue with Load Balancer health checks and IPVS. (kube-proxy) Merged PR90260 to fix the missing network monitoring issue for containerd clusters. (kubelet) Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. (kubelet) Merged PR72914 to fix the issue where immediately creating and scheduling a Pod to the same node after its deletion could cause volume mount failures. (kube-controller-manager) Fixed the issue where creating containers on CentOS could cause cgroup leaks. (kubelet) Merged PR98262 to support dynamic log level adjustment for kube-controller-manager. (kube-controller-manager). Merged PR97752 to fix the issue where NewReplicaSet is displayed as <none> when a deployment is described. (kubectl) Merged PR94833 to fix the issue where the image tag in the status does not match when a Pod image has multiple tags. (kubelet). Merged PR100060 to automatically delete volume directories left behind by orphaned Pods. (kubelet). kube-controller-manager supports super nodes. (kube-controller-manager). kube-scheduler supports retaining a fixed number of local replicas when a hybrid cloud is scaled out to TKE Serverless. (kube-scheduler) Added support for cbs csi migration. (kube-controller-manager, kubelet). Merged PR93260 to fix the issue where the AWS Credential Provider caused slow node startup. (kubelet). Added the command-line argument eks-config-namespace to the scheduler to specify the namespace containing the configuration for scaling out TKE Serverless. (kube-scheduler) TKE supports hybrid cloud nodes. (kube-controller-manager). |
Date | Version | Update |
2026-02-26 | v1.18.4-tke.60 | Fixed the issue where the kube-scheduler topology feature did not behave as expected. (kube-scheduler). |
2026-01-27 | v1.18.4-tke.59 | Added the DisableFdSocketMetrics feature gate (disabled by default). When enabled, it prevents the collection of fd and socket information. (kubelet). |
2025-12-17 | v1.18.4-tke.58 | Fixed the issue where the binary version number did not match the image tag. (kubelet, kube-apiserver, kube-scheduler, kube-controller-manager). |
2025-08-07 | v1.18.4-tke.57 | Optimized in-place kubelet configuration changes. (kubelet) Fixed the issue of failing to schedule monthly subscription super nodes. (kube-scheduler) |
2025-05-15 | v1.18.4-tke.56 | Optimized independent apiserver. (kube-apiserver) |
2025-05-12 | v1.18.4-tke.55 | PR 99208: Fixed the issue where a saveVolumeData failure could potentially delete business data. (kubelet). Supported independent apiserver. (kube-apiserver) Mitigated apiserver cache breakdown. (kube-apiserver) |
2025-03-25 | v1.18.4-tke.54 | PR112607: Fixed the issue where the automatic binding of the kubelet directory to the data disk resulted in residual CBS mount points. (kubelet) |
2025-01-09 | v1.18.4-tke.53 | Fixed in-place specification adjustment: Fixed the bug that repeatedly calling kubelet admit might trigger the device manager reuse map to be empty, causing kubelet panic. (kubelet) Allowed Pods to add or delete CPU or memory resources while ensuring Pod QOSClass remained unchanged. Allowed Pods to reduce limits. (apiserver) Optimized resource awareness policies. (kube-scheduler) |
2024-12-20 | v1.18.4-tke.52 | Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-11-27 | v1.18.4-tke.51 | Fixed the issue where static PV Pods could not be scheduled when the cluster had chc nodes. (kube-scheduler) |
2024-11-03 | v1.18.4-tke.50 | PR102498: Optimized scheduling logic to skip non-existent Pods. (kube-scheduler) |
2024-09-20 | v1.18.4-tke.49 | Changed the validity period of self-signed certificates from 1 year to 5 years. (kube-apiserver) Governed multi-backend risks with iptables in early versions of kube-proxy. (kube-proxy). Optimized scheduling logic. (kube-scheduler) Supported in-place Pod specification upgrades. (kube-apiserver/kube-scheduler/kubelet) |
2024-07-16 | v1.18.4-tke.48 | Fixed the issue of extra suffixes in the release version. |
2024-07-09 | v1.18.4-tke.45 | Merged PR89808 to fix the issue where kubelet 1.18 might cause goroutine leakage when a large number of exec operations were executed. (kubelet) Merged PR93250 to fix the type conversion issue. (kube-apiserver) Enabled kubelet to support Pod deletion protection. (kubelet) |
2024-04-10 | v1.18.4-tke.44 | Optimized PVC mounting. (kube-scheduler). |
2024-02-27 | v1.18.4-tke.43 | Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler). Optimized the compute resource policy for super nodes. (kube-scheduler) |
2024-02-01 | v1.18.4-tke.42 | Native nodes support CVM mode. (kubelet). Merged PR102014 to fix the issue where CPUs bound to init containers became unusable after release. (kubelet). |
2023-12-04 | v1.18.4-tke.41 | CVE-2022-3294 (kube-apiserver) Implemented overload protection for the apiserver. (kube-apiserver). Added annotation support for mounting cgroupfs to Pods. (kubelet). Added userAgent reporting to metrics and detailed logging when resources are deleted. (kube-apiserver) |
2023-07-20 | v1.18.4-tke.39 | Enabled scheduling of daemonset Pods to monthly subscription super nodes. (kube-scheduler). Backported PR107831 to include pod UID in the kubelet secret/configmap cache. (kubelet). Fixed the issue of abnormal disk mounting for Pods. (kube-controller-manager/kubelet). |
2023-05-26 | v1.18.4-tke.37 | Improved the handling logic for Pod in-place resizing. (kube-apiserver, kube-scheduler, kubelet) |
2023-05-16 | v1.18.4-tke.36 | During pod garbage collection (gc), kube-controller-manager prioritizes the deletion of Evicted pods. (kube-controller-manager) Fixed the issue where GET requests were not logged as WATCH operations under certain circumstances. (kube-apiserver) |
2023-03-03 | v1.18.4-tke.35 | Added a delegated_to_storage Tag to the list_too_many_objects_events_total metric to indicate whether the request is delegated to backend storage. (kube-apiserver)Backported PR89160 to delete symbolic links associated with a container only after the container has completely exited. (kubelet) Added support for the exec/logs commands on super nodes and DaemonSet Pods. (kube-apiserver). Optimized super node scheduling and added a unified EnableSuperNodeFeatures switch. (kube-scheduler)Fixed the issue where inotify leaks cause kubelet memory leaks. (kubelet) Removed logs related to watch_too_old_objects_events_total, list_too_many_objects_events_total, and watch_too_many_objects_events_total to avoid excessive information output. (kube-apiserver)Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) Fixed the issue where the rest_client_rate_limiter_duration_seconds metric was not registered. (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, kube-proxy)Merged PR110951 to prevent the issue where a PV mount operation occasionally hangs, causing all Pods associated with that PV to be stuck in the Creating state. (kubelet) Merged PR111635 to fix the issue where ipvs real server weights become zero after kube-proxy restarts. (kube-proxy) |
2023-01-05 | v1.18.4-tke.34 | Super nodes support running DaemonSet Pods when needed. (kube-controller-manager). Merged PR102576 to stop using RemoveAll when leftover Pod volumes are deleted, preventing the deletion of volume-related data. (kubelet) In-place updates support reducing resource limits. (kube-apiserver). Fixed the issue where terminated containers were not deleted during log rotation. (kubelet) Records metrics when list/watch operations encounter too old, too large, or too many objects (kube-apiserver): watch_too_old_objects_events_total, list_too_large_objects_events_total, list_too_many_objects_events_total, watch_too_many_objects_events_total |
2022-11-09 | v1.18.4-tke.33 | Optimized the scheduling of yearly/monthly subscription super nodes. (kube-scheduler) Optimized the scheduling of local replica counts for super nodes. |
2022-10-13 | v1.18.4-tke.30 | Merged PR112299 to optimize the gzip compression level for the apiserver. (kube-apiserver). Fixed CVE-2022-3172. (kube-apiserver) Added the calling_webhook_timeout_error metric. (kube-apiserver). Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). Added the StatusDetaching state for CBS disks. (kube-controller-manager) Serverless clusters support DaemonSets. (kube-apiserver). |
2022-09-07 | v1.18.4-tke.28 | Optimized List performance for large-scale clusters. (kube-apiserver). Optimized super node scheduling. (kube-scheduler). Supported scheduling Pods in the kube-system namespace to prepaid super nodes. (kube-scheduler). Optimized HPA for super nodes. (kube-controller-manager). Supported in-place updates for Pod resources. (kube-apiserver,kubelet). Merged PR97348 to fix the issue where HPA scaling quantities were incorrect when StabilizationWindowSeconds was set. (kube-controller-manager). Optimized the scheduling of yearly/monthly subscription super nodes. Merged PR108831 to fix the kubelet panic issue caused by creating multiple Pods simultaneously. (kubelet). |
2022-07-27 | v1.18.4-tke.26 | CBS supports native nodes. (kubelet) |
2022-03-18 | v1.18.4-tke.23 | Merged PR92878 to print warning messages only when the setting of ConfigMap/Secret volume Ownership exceeds 30 seconds, thereby avoiding excessive log generation. (kubelet) Merged PR106906 to detect whether the subpath of the network storage volume has been deleted, thereby preventing Pods from remaining in the terminating state. (kubelet) When anti-affinity scheduling is performed based on hostname, TKE Serverless super nodes are ignored. (kube-scheduler) Merged PR93026 to fix the issue where DefaultPodTopologySpread could not obtain replicaset information. (kube-scheduler) |
2022-01-20 | v1.18.4-tke.20 | TKE Serverless rescheduling optimization: Reduces the score for super nodes within the same AZ that have been evicted. (kube-scheduler) The apiserver supports integrating external services of the ExternalName 556 type. (kube-apiserver) Added support for binding Load Balancer (LB) addresses to the ipvs network interface. (kube-proxy) |
2021-12-09 | v1.18.4-tke.17 | Resolved the issue where kube-controller-manager was rate-limited when accessing the api-server due to a large number of volume attachment objects in the cluster. (kube-controller-manager) Merged PR95650 to have HPA ignore deleted Pods when the number of replicas is calculated. (kube-controller-manager) Fixed the issue where TKE Serverless cpu resource calculations were inconsistent with the frontend. (kube-scheduler) |
2021-12-02 | v1.18.4-tke.16 | Fixed the bug that occurs when scheduling to super nodes is performed. (kube-scheduler) Optimized the super node scheduling algorithm. (kube-scheduler). |
2021-11-26 | v1.18.4-tke.15 | Merged PR96444 to retry on errors encountered during RBAC policy synchronization. (kube-apiserver) Added support for custom installation of other CNIs on hybrid cloud external nodes. (kube-controller-manager) Added support for the requirement of core binding for Cloud Gaming Android container groups. (kubelet) Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Merged PR99336 to improve the synchronization mechanism for node information during kubelet startup. (kubelet) Fixed CVE-2021-25741 to prevent unauthorized access to host files via symbolic links. (kubelet) Optimized the error message for scheduling timeouts caused by cbs disk creation failures. (kube-scheduler) Optimized grpc logs to prevent excessive logging when kubelet collects volume status. (kubelet) Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) |
2021-08-23 | v1.18.4-tke.14 | When a TKE cluster is scaled out to TKE Serverless, the scheduler supports fixed IPs. (kube-scheduler) When a TKE cluster is scaled out to TKE Serverless, the scheduler skips other pre-selection policies upon matching a TKE Serverless fixed IP address. (kube-scheduler) When a TKE cluster is scaled out to TKE Serverless, the scheduler optimizes scheduling for TKE Serverless nodes, including resource-aware rescheduling, priority model scheduling, and optimized pre-selection/selection policies for TKE Serverless nodes. (kube-scheduler) Logged the loaded ipvs kernel modules to prevent kube-proxy crashes in ipvs mode. (kube-proxy) Prevented panic when an error occurs while writing the cpu manager state file. (kubelet) |
2021-07-22 | v1.18.4-tke.13 | Merged PR91859, fixing the issue where kube-apiserver panics when a CRD type has only one letter. (kube-apiserver) |
2021-07-13 | v1.18.4-tke.12 | When a scale-out from a TKE cluster to EKS is performed, the scheduler correctly perceives the remaining IPs in the subnet during batch Pod scheduling, and schedules the appropriate number of Pods to super nodes. (kube-scheduler) Added support for collecting disk usage metrics for the Containerd runtime. (kubelet) Added support for specifying Pods during scale-in. (kube-controller-manager) |
2021-06-05 | v1.18.4-tke.11 | TKE supports hybrid cloud nodes. (kube-controller-manager) |
2021-05-14 | v1.18.4-tke.9 | Backported PR93370 to add support for CronJobControllerV2. (kube-controller-manager) Merged PR100376 to enable HTTP/2 health checks, preventing issues where connections cannot be recovered after being lost. (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, kube-proxy, kubectl) Merged PR100317 to fix the issue where node updates could bypass the Validating Admission Webhook in CVE-2021-25735. (kube-apiserver) When a scale-out from a TKE cluster to TKE Serverless is performed, the platform supports ComputeResource, TKE Serverless ClusterIP, and HPA. (kube-controller-manager, kube-scheduler) |
2021-04-02 | v1.18.4-tke.8 | Merged PR97752 to fix the issue where NewReplicaSet is displayed as <none> when a deployment is described. (kubectl) Merged PR93808 to fix the issue where executing `kube-scheduler --version` returns extraneous information. (kube-scheduler) Merged PR91590 to fix the issue where a warning about ports already being allocated is displayed when a NodePort-type multi-protocol Service is used. (kube-apiserver) Merged PR98262 to support dynamic log level adjustment for kube-controller-manager. (kube-controller-manager). Merged PR95154 to fix the issue where the kube-scheduler snapshot includes nodes that are being deleted. (kube-scheduler). Merged PR95711 to fix the issue where the `kubectl drain` command consumes high CPU. (kubectl). Merged PR96602 to fix the memory leak issue in the apiserver when time jumps forward or backward. (kube-apiserver). Merged PR97023 to delete the relevant metadata directories when emptyDir-type volumes are unmounted. (kubelet) Merged PR97527 to fix the issue of unsynchronized map access operations in the cpumanager. (kubelet). Merged PR100190 to automatically delete volume directories left behind by orphaned Pods. (kubelet). Merged PR92614. When all containers of a Pod with a restart policy of RestartPolicyOnFailure exit successfully, a new Sandbox is no longer created. (kubelet) Merged PR94833 to fix the issue where the image tag in the status does not match when a Pod image has multiple tags. (kubelet). |
2020-12-28 | v1.18.4-tke.6 (ARM cluster support starts from this version). | Added metrics for QcloudCbs. (kube-controller-manager) Fixed the issue of extra spaces when the serial value is viewed while a cbs disk is mounted. (kubelet) |
2020-12-21 | v1.18.4-tke.5 | Merged PR94712 to fix CVE-2020-8564 - Docker configuration leaks when the file format is incorrect and the logLevel is >= 4. (kubelet). Merged PR95316 to fix CVE-2020-8565 - a log token leak caused by an incomplete fix for CVE-2019 -11250. (logLevel >= 9) (kube-apiserver, kubectl). Merged PR95245 to fix CVE-2020-8566 - Ceph RBD adminSecrets are exposed in logs when the loglevel is >= 4. (kube-controller-manager). Fixed the issue where restarting kubelet causes Pod readiness checks to fail. (kubelet). Merged PR90825 to resolve the issue where a race condition could cause the Pop operation of the fifo queue in client-go to hang, resulting in pods remaining in a pending state. (kubelet). The scheduler supports super nodes. (kube-scheduler) kube-controller-manager supports super nodes. (kube-controller-manager). Sets the instance-type Tag based on the node's actual machine model, instead of fixing it to QCLOUD. (kubelet). Added the CBS section to OpenAPI. (kube-apiserver). Merged PR91126 to fix the issue where the scheduler cache becomes inconsistent when Pods have the same name but different UIDs. (kube-scheduler). Merged PR93387 to fix the issue where corrupted node cache information in the scheduler prevents daemonset pods from being scheduled to certain nodes. (kube-scheduler). Merged PR89465 to fix the issue where HPA, based on pod metrics, incorrectly calculates the number of instances during rolling updates. (kube-controller-manager). |
2020-10-13 | v1.18.4-tke.3 | Merged PR89629 to resolve the issue where containers mounting subpaths fail to restart permanently after a configmap change. (kubelet). QcloudCbs supports BulkVolumeVerification. (kube-controller-manager) Merged PR94430 to fix the issue where the client-go reflector fails to detect the "Too large resource version" error. (kubelet). |
2020-08-12 | v1.18.4-tke.2 | Merged PR93403, removing the erroneous log messages printed when kubelet updates Pod Conditions that do not belong to it. (kubelet) |
2020-08-04 | v1.18.4-tke.1 | Reverted PR63066 to fix the issue with Load Balancer health checks and IPVS. (kube-proxy) Merged PR72914 to fix the issue where immediately creating and scheduling a Pod to the same node after its deletion could cause volume mount failures. (kube-controller-manager) Fixed the issue where creating containers on CentOS could cause cgroup leaks. (kubelet) Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. (kubelet) Added caching and timeout to metadata. Added support for using node names as hostnames to cloud-provider. (kubelet) Added local caching to metadata. (kubelet) Merged CBS and related fix code. (kubelet) Merged PR90260 to fix the missing network monitoring issue for containerd clusters. (kubelet) TKE supports detecting the maximum number of qcloudcbs volumes that can be mounted on a single node. For version 1.12 and later, this limit is maxAttachCount-2, while for version 1.10, it defaults to 18. (kube-scheduler) CBS intree resolves the issue where disks continue to be unmounted when they do not exist, which generates a large number of invalid requests. (kubelet). Merged PR2359 to fix the monitoring gap issue caused by the inability to obtain the docker root directory. (kubelet). kube-scheduler supports dynamic log level configuration. (kube-scheduler). Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for CBS is missing, enabling users to use CBS normally. (kubelet) TKE detects the maximum number of qcloudcbs volumes that can be mounted on a single node, while the kubelet side does not patch the node. (kubelet). Merged PR89296 to stop logging whether the iptables random-fully parameter is enabled. (kube-proxy). Fixed the aws issue, PR92162. (kubelet) Merged PR91277 to prevent the issue where CLB health checks cause kube-apiserver to generate a large number of TLS handshake error logs. (kube-apiserver) Merged PR91500 to fix the issue where the KUBERNETES_SERVICE_HOST environment variable is missing. (kubelet) Merged PR92537 to fix the issue where the client-go reflector fails to recover from the "Too large resource version" error. (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, and kube-proxy). Merged PR92969 to fix CVE-2020-8559, which allows privilege escalation from a compromised node to infiltrate other nodes. (kube-apiserver). Merged PR92921 to fix CVE-2020-8557, a DOS attack issue that exhausts disk space by writing to "/etc/hosts". (kubelet). |
Date | Version | Update |
2026-03-24 | v1.16.3-tke.46 | Fixed the issue where the kube-scheduler topology feature did not behave as expected. (kube-scheduler). |
2025-08-07 | v1.16.3-tke.45 | Optimized in-place kubelet configuration changes. (kubelet) |
2025-05-15 | v1.16.3-tke.44 | Converged policies related to super nodes. (kube-scheduler) Fixed the eks daemonset injection issue. (kube-controller-manager) PR99208: Fixed the issue where a saveVolumeData failure might delete business data. (kubelet) |
2025-01-09 | v1.16.3-tke.43 | Upgraded structured-merge-diff to version 1.0.1 to address performance issues with large CRD structures. (kube-apiserver) Fixed in-place specification adjustment: Fixed the bug that repeatedly calling kubelet admit might trigger the device manager reuse map to be empty, causing kubelet panic. (kubelet) Allowed Pods to add or delete CPU or memory resources while ensuring Pod QOSClass remained unchanged. Allowed Pods to reduce limits. (apiserver) |
2024-12-20 | v1.16.3-tke.42 | Changed the validity period of self-signed certificates from 1 year to 5 years. (kube-apiserver) Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) Supported the reverse probing node of kcm. (kube-controller-manager) |
2024-07-09 | v1.16.3-tke.41 | Implemented etcd overload protection. (kube-apiserver) |
2024-04-10 | v1.16.3-tke.40 | Native nodes support cvm mode. (kubelet). Optimized volume reconciliation. (kube-controller-manager) Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler) Fixed the scheduler cache issue. (kube-scheduler). Optimized PVC mounting. (kube-scheduler). |
2023-03-24 | v1.16.3-tke.35 | Optimized the installation script. (kubelet). |
2023-03-03 | v1.16.3-tke.34 | Super nodes support running DaemonSet Pods when needed. (kube-controller-manager). Merged PR102576 to stop using RemoveAll when leftover Pod volumes are deleted, preventing the deletion of volume-related data. (kubelet) In-place updates support reducing resource limits. (kube-apiserver). Added support for the exec/logs commands on super nodes and DaemonSet Pods. (kube-apiserver). Fixed the issue where inotify leaks cause kubelet memory leaks. (kubelet). Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) Merged PR111635 to fix the issue where ipvs real server weights become zero after kube-proxy restarts. (kube-proxy). |
2022-10-13 | v1.16.3-tke.32 | Merged PR112299 to optimize the gzip compression level for the apiserver. (kube-apiserver). Fixed CVE-2022-3172. (kube-apiserver) Added the calling_webhook_timeout_error metric. (kube-apiserver). Ignore errors caused by TKEDefaultQuota conflicts and increase the number of workers to resolve the resource quota evaluates timeout error. (kube-apiserver). When processing Pods, CPU Manager first cleans up terminated containers to prevent CPU allocation failures. (kubelet) Added the StatusDetaching state for CBS disks. (kube-controller-manager) Serverless clusters support DaemonSets. (kube-apiserver). Supported in-place updates for Pod resources. (kube-apiserver,kubelet). |
2022-07-27 | v1.16.3-tke.28 | When anti-affinity scheduling is performed based on hostname, TKE Serverless virtual nodes are ignored. (kube-scheduler) TKE Serverless supports the sandbox retention feature. (kube-scheduler). CBS supports native nodes. (kubelet). |
2022-03-18 | v1.16.3-tke.27 | Supports specifying pods during scale-in. (kube-controller-manager) Optimized the super node scheduling algorithm. (kube-scheduler). |
2022-01-20 | v1.16.3-tke.25 | kube-proxy supports binding Load Balancer (LB) addresses to the ipvs network interface. The apiserver supports integrating external services of the ExternalName type. (kube-apiserver) Optimized TKE Serverless scheduling. (kube-scheduler). |
2021-12-09 | v1.16.3-tke.24 | Fixed the issue where the TKE Serverless local replica quantity policy failed on statefulset Pods. (kube-scheduler) |
2021-12-02 | v1.16.3-tke.23 | Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Fixed the bug that occurs when scheduling to super nodes is performed. (kube-scheduler) |
2021-09-03 | v1.16.3-tke.22 | Avoid panic when an error occurs while writing the cpu manager state file. (kubelet) |
2021-08-17 | v1.16.3-tke.21 | Optimized scheduling for TKE Serverless nodes, including resource-aware rescheduling, priority model scheduling, and optimized pre-selection/selection policies for TKE Serverless nodes. (kube-scheduler) Backported 87692 to fix the issue where the scheduler's pending_pods and schedule_attempts_total metrics had no data. (kube-scheduler) |
2021-07-19 | v1.16.3-tke.20 | When a TKE cluster is scaled out to TKE Serverless, the scheduler correctly perceives the remaining IPs in the subnet during batch Pod scheduling and schedules the appropriate number of Pods to super nodes. (kube-scheduler) Merged PR88507 to resolve the inconsistency between podIP and podIPs when Pod status is updated. (kube-apiserver) |
2021-05-24 | v1.16.3-tke.17 | Backported PR93370 to add support for CronJobControllerV2. (kube-controller-manager) When a scale-out from a TKE cluster to TKE Serverless is performed, the scheduler supports retaining the number of local replicas. (kube-scheduler) |
2021-05-06 | v1.16.3-tke.16 | Updated the startup method for kube-proxy when it is running in image mode to automatically adapt to the iptables mode of the node, supporting operating systems that run iptables in nf_tables mode by default. |
2021-04-14 | v1.16.3-tke.15 | Merged PR97752 to fix the issue where NewReplicaSet is displayed as <none> when a deployment is described. (kubectl) Merged PR92614. When all containers of a Pod with a restart policy of RestartPolicyOnFailure exit successfully, a new Sandbox is no longer created. (kubelet) Merged PR91590 to fix the issue where a warning about ports already being allocated is displayed when a NodePort-type multi-protocol Service is used. (kube-apiserver) Merged PR98262 to support dynamic log level adjustment for kube-controller-manager. (kube-controller-manager). Merged PR95301 to automatically delete volume directories left behind by orphaned Pods. (kubelet). |
2020-12-28 | v1.16.3-tke.14 | Added metrics for QcloudCbs. (kube-controller-manager) Fixed the issue of extra spaces when the serial value is viewed while a cbs disk is mounted. (kubelet) |
2020-12-21 | v1.16.3-tke.13 | Merged PR94712 to fix CVE-2020-8564 - Docker configuration leaks when the file format is incorrect and the logLevel is >= 4. (kubelet). Merged PR95316 to fix CVE-2020-8565 - a log token leak caused by an incomplete fix for CVE-2019-11250 when the logLevel >= 9. (kube-apiserver, kubectl). Merged PR95245 to fix CVE-2020-8566 - Ceph RBD adminSecrets are exposed in logs when the loglevel is >= 4. (kube-controller-manager). Merged PR86191 to fix the issue where Pods might enter an erroneous state during node restarts. (kubelet). Merged PR86140 to fix the issue where the controller manager's failure to properly handle timeout errors prevented scale-out Pods from being created. (kube-controller-manager). Merged PR90825 to resolve the issue where a race condition could cause the Pop operation of the fifo queue in client-go to hang, resulting in pods remaining in a pending state. (kubelet). The scheduler supports super nodes. (kube-scheduler) kube-controller-manager supports super nodes. (kube-controller-manager). Sets the instance-type Tag based on the node's actual machine model, instead of fixing it to QCLOUD. (kubelet). Added the CBS section to OpenAPI. (kube-apiserver). Merged PR81344 to fix the issue where the CPU Manager does not support SourcesReady. (kubelet). Merged PR91126 to fix the issue where the scheduler cache becomes inconsistent when Pods have the same name but different UIDs. (kube-scheduler). Merged PR89224 to fix the issue where the lack of checks in NodeInfo caused abnormal restarts of kube-scheduler. (kube-scheduler). Merged PR89465 to fix the issue where HPA, based on pod metrics, incorrectly calculates the number of instances during rolling updates. (kube-controller-manager). |
2020-10-13 | v1.16.3-tke.11 | Merged PR92971 to fix CVE-2020-8559, which allows privilege escalation from a compromised node to infiltrate other nodes. (kube-apiserver). Merged PR92924 to fix CVE-2020-8557, a DOS attack issue that exhausts disk space by writing to /etc/hosts. (kubelet). Merged PR93403 to remove the erroneous log message stating that kubelet's update does not belong to kubelet's Pod Condition. (kubelet). Merged PR89629 to resolve the issue where containers mounting subpaths fail to restart permanently after a configmap change. (kubelet). QcloudCbs supports BulkVolumeVerification. (kube-controller-manager) Merged PR84998 to resolve the issue where the corresponding node lease object might be recreated after Node deletion, leading to garbage data. (kubelet). |
2020-07-28 | v1.16.3-tke.10 | Merged PR91277 to prevent the issue where CLB health checks cause kube-apiserver to generate a large number of TLS handshake error logs. (kube-apiserver) Merged PR91500 to fix the issue where the KUBERNETES_SERVICE_HOST environment variable is missing. (kubelet) |
2020-06-17 | v1.16.3-tke.9 | Temporarily fixed the AWS issue PR92162. No longer registers the AWS Credential Provider to avoid the node startup slowdown issue caused by it. |
2020-06-11 | v1.16.3-tke.8 | |
2020-06-10 | v1.16.3-tke.7 | Merged PR90260 to fix the missing network monitoring issue for containerd clusters. Merged PR89515 to fix the issue where HPA incorrectly calculates the number of instances during rolling updates. Merged PR91252 to ignore Pod Condition updates generated by other components, thereby avoiding unnecessary scheduling. Merged PR89794 to clean up error logs in kube-controller-manager, thereby preventing the Half-Blind SSRF attack associated with CVE-2020-8555. |
2020-05-18 | v1.16.3-tke.6 | tke can perceive the maximum number of qcloudcbs that can be mounted on a single Node, but does not support dynamically obtaining the maximum value. |
2020-04-20 | v1.16.3-tke.5 | Merged PR69047 to resolve the backward compatibility issue with node.Spec.Unschedulable (this fix was overwritten when the in-tree cbs code was merged). |
2020-04-14 | v1.16.3-tke.4 | Merged PR87913 to fix CVE-2020-8551: the Kubelet DoS attack issue. Merged PR87669 to fix CVE-2020-8552: the kube-apiserver DoS attack issue. TKE supports detecting the maximum number of qcloudcbs volumes that can be mounted on a single Node (for version 1.12 and later, this limit is maxAttachCount-2, while for version 1.10, it defaults to 18). Merged PR87467 to fix the issue where an authorized user sending malicious YAML causes kubectl to consume excessive CPU when parsing YAML. |
2020-03-11 | v1.16.3-tke.3 | cbs intree resolves the issue where disks continue to be unmounted when they do not exist, which generates a large number of invalid requests. Added local caching to metadata. |
2020-02-14 | v1.16.3-tke.2 | Merged PR2359 to fix the monitoring gap issue caused by the inability to obtain the docker root directory. Merged PR86583 to increase the log level when iptables does not support random-fully, thereby avoiding excessive log generation. kube-scheduler supports dynamic log level configuration. Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for cbs is missing, enabling users to use cbs normally. Merged PR86230 to skip updating the scheduling of assumed pods during the pod scheduling process. |
2020-01-06 | v1.16.3-tke.1 | Merged PR79036 to fix the issue where CPU quota is disabled when the CPU Manager is enabled and a pod's QoS is Guaranteed. Merged PR84167 to fix the issue where apiserver health checks fail due to incorrect ETCD key prefixes. Reverted PR63066 to fix the issue with CLB health checks and IPVS. Merged PR72914 to fix the issue where immediately creating and scheduling a Pod to the same node after its deletion could cause volume mount failures. Fixed the issue where creating containers on CentOS could cause cgroup leaks. Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. Added caching and timeout to metadata. Added support for using node names as hostnames to cloud-provider. Reverted PR79036 to fix the issue where CPU quota is disabled when the CPU Manager is enabled and a pod's QoS is Guaranteed. Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for cbs is missing, enabling users to use cbs normally. |
Date | Version | Update |
2025-05-15 | v1.14.3-tke.34 | Converged policies related to super nodes. (kube-scheduler) |
2025-01-09 | v1.14.3-tke.33 | Changed the validity period of self-signed certificates from 1 year to 5 years. (kube-apiserver) Governed multi-backend risks with iptables in earlier versions of kube-proxy. (kube-proxy) |
2024-07-09 | v1.14.3-tke.32 | Implemented etcd overload protection. (kube-apiserver) |
2024-04-10 | v1.14.3-tke.31 | Optimized volume reconciliation. (kube-controller-manager) Super nodes allow scheduling of Pods in TKE trunking ENI mode. (kube-scheduler) |
2023-03-03 | v1.14.3-tke.28 | Added the StatusDetaching state for CBS disks. (kube-controller-manager) Fixed CVE-2022-3172. (kube-apiserver) Migrated 16+ webhook metrics and added the calling_webhook_timeout_error metric. (kube-apiserver)Migrated the etcd_request_duration_seconds metric. (kube-apiserver)Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) Merged PR111635 to fix the issue where ipvs real server weights become zero after kube-proxy restarts. (kube-proxy) |
2022-04-13 | v1.14.3-tke.27 | Merged PR78428 to avoid writing warning messages when iptables rules are exported, which could cause kube-proxy to panic during recovery. (kube-proxy) |
2022-03-18 | v1.14.3-tke.26 | Supports specifying pods during scale-in. (kube-controller-manager) Optimized the super node scheduling algorithm. (kube-scheduler) Merged PR80851 to fix CVE-2019-11247 and prevent unauthorized access to CRD resources. (kube-apiserver) |
2022-01-20 | v1.14.3-tke.24 | kube-proxy supports binding Load Balancer (LB) addresses to the ipvs network interface. The apiserver supports integrating external services of the ExternalName type. (kube-apiserver) Optimized TKE Serverless scheduling. (kube-scheduler). |
2021-12-02 | v1.14.3-tke.23 | When a scale-out from a TKE cluster to TKE Serverless is performed, the scheduler correctly perceives the remaining IPs in the subnet during batch Pod scheduling and schedules the appropriate number of Pods to super nodes. (kube-scheduler) Optimized scheduling for TKE Serverless nodes: resource-aware rescheduling, priority model scheduling, and optimized pre-selection/selection policies for TKE Serverless nodes. (kube-scheduler). Added support for the Prebind and Unreserve operations of the extended scheduler. (kube-scheduler). Prevented Pods using Cloud Block Storage (CBS) from being scheduled to external CHC nodes. (kube-scheduler) Fixed the bug that occurs when scheduling to super nodes is performed. (kube-scheduler) |
2021-05-06 | v1.14.3-tke.22 | Updated the startup method for kube-proxy when it is running in image mode to automatically adapt to the iptables mode of the node, supporting operating systems that run iptables in nf_tables mode by default. |
2021-04-14 | v1.14.3-tke.21 | Merged PR97752 to fix the issue where NewReplicaSet is displayed as <none> when a deployment is described. (kubectl) Merged PR78999 to fix the case-sensitivity issue in protocol detection during graceful shutdown. (kube-proxy) Merged PR91590 to fix the issue where a warning about ports already being allocated is displayed when a NodePort-type multi-protocol Service is used. (kube-apiserver) Merged PR98262 to support dynamic log level adjustment for kube-controller-manager. (kube-controller-manager). Merged PR95301 to automatically delete volume directories left behind by orphaned Pods. (kubelet). |
2020-12-28 | v1.14.3-tke.19 | Added metrics for QcloudCbs. (kube-controller-manager) Fixed the issue of extra spaces when the serial value is viewed while a cbs disk is mounted. (kubelet) |
2020-12-21 | v1.14.3-tke.18 | Merged PR94712 to fix CVE-2020-8564 - Docker configuration leaks when the file format is incorrect and the logLevel is >= 4. (kubelet). Merged PR95316 to fix CVE-2020-8565 - a log token leak caused by an incomplete fix for CVE-2019-11250 when the logLevel >= 9. (kube-apiserver, kubectl). Merged PR95245 to fix CVE-2020-8566 - Ceph RBD adminSecrets are exposed in logs when the loglevel is >= 4. (kube-controller-manager). Merged PR86140 to fix the issue where the controller manager's failure to properly handle timeout errors prevented scale-out Pods from being created. (kube-controller-manager). The scheduler supports super nodes. (kube-scheduler) kube-controller-manager supports super nodes. (kube-controller-manager). Sets the instance-type Tag based on the node's actual machine model, instead of fixing it to QCLOUD. (kubelet). Merged PR79338 to disable the pids cgroup subsystem when both SupportPodPidsLimit and SupportNodePidsLimit are not enabled. (kubelet). Merged PR89224 to fix the issue where the lack of checks in NodeInfo caused abnormal restarts of kube-scheduler. (kube-scheduler). Merged PR89465 to fix the issue where HPA, based on pod metrics, incorrectly calculates the number of instances during rolling updates. (kube-controller-manager). |
2020-10-13 | v1.14.3-tke.17 | Merged PR74781 to change the default update policy for ConfigMap and Secret from Cache to Watch. (kubelet) Merged PR93403 to remove the erroneous log message for Pod Conditions that do not belong to kubelet during its updates. (kubelet). Merged PR89629 to resolve the issue where containers mounting subpaths fail to restart permanently after a configmap change. (kubelet). Merged PR80942 to fix the issue where rules are not deleted after a service is removed in ipvs mode. (kube-proxy). QcloudCbs supports BulkVolumeVerification. (kube-controller-manager) |
2020-08-04 | v1.14.3-tke.16 | Merged PR78883, fixing the bug where a default value was incorrectly added to pod.spec.container.SecurityContext.ProcMount. |
2020-07-28 | v1.14.3-tke.15 | Merged PR76518 and PR82514 to limit the response size of http and exec probes, preventing them from occupying a large amount of node memory. (kubelet) Merged PR91277 to prevent the issue where CLB health checks cause kube-apiserver to generate a large number of TLS handshake error logs. (kube-apiserver) Merged PR91500 to fix the issue where the KUBERNETES_SERVICE_HOST environment variable is missing. (kubelet) Merged PR77475 to fix the issue where Cronjob cannot be scheduled when the number of Jobs exceeds 500. (kube-controller-manager). |
2020-06-10 | v1.14.3-tke.14 | Merged PR85027 to fix the issue where HPA incorrectly calculates the number of instances during rolling updates. Merged <|diff_insert|>PR79708 to use spec.replicas for calculating the current number of replicas for HPA. Merged PR91252 to ignore Pod Condition updates generated by other components, thereby avoiding unnecessary scheduling. Merged PR89794 to clean up error logs in kube-controller-manager, thereby preventing the Half-Blind SSRF attack associated with CVE-2020-8555. |
2020-06-04 | v1.14.3-tke.13 | Merged PR90260 to fix the missing network monitoring issue for containerd clusters. Merged PR79451 to fix the issue where kubelet does not recreate a SandBox after a creation failure when restartPolicy is Never. |
2020-05-18 | v1.14.3-tke.12 | tke can perceive the maximum number of qcloudcbs that can be mounted on a single node, but does not support dynamically obtaining the maximum value. |
2020-04-14 | v1.14.3-tke.11 | Merged PR75442 to correct the bandwidth unit from Kb to b. Merged PR87669 to fix CVE-2020-8552: the apiserver DoS attack issue. tke supports detecting the maximum number of qcloudcbs volumes that can be mounted on a single node (for version 1.12 and later, this limit is maxAttachCount-2, while for version 1.10, it defaults to 18). |
2020-04-14 | v1.14.3-tke.10 | cbs intree resolved the issue where attempting to unmount a disk when it does not exist generates a large number of invalid requests. |
2020-01-13 | v1.14.3-tke.9 | Merged PR2359 to fix the monitoring gap issue caused by the inability to obtain the docker root directory. Merged PR86583 to increase the log level when iptables does not support random-fully, thereby avoiding excessive log generation. kube-scheduler supports dynamic log level configuration. Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for cbs is missing, enabling users to use cbs normally. Merged PR86230 to skip updating the scheduling of assumed pods during the pod scheduling process. |
2019-12-23 | v1.14.3-tke.8 | Reverted PR79036 to resolve the issue where cpu quota was disabled when cpumanager was enabled and the Pod's QoS was Guaranteed. |
2019-12-17 | v1.14.3-tke.7 | Added caching and timeout to metadata. Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. Prevent Pods from becoming not ready due to readiness probe failures during kubelet restarts. |
2019-11-28 | v1.14.3-tke.6 | cloud-provider added support for using the node name as the hostname. |
2019-11-18 | v1.14.3-tke.5 | Merged PR83435 to fix the issue where an attacker could send specially crafted malicious YAML or JSON payloads, causing kube-apiserver to exhaust CPU or memory resources and become unavailable. Merged PR84167 to fix the issue where apiserver health checks fail due to incorrect Etcd key prefixes. Merged PR75622 to fix the issue where changes to StatefulSets (sts) experience high latency (approximately 20 seconds) when the changes are synchronized to Pods in clusters with a large number of sts workloads (>2000). |
2019-10-23 | v1.14.3-tke.4 | Merged PR79036 to resolve the issue where cpu quota was disabled when CPU Manager was enabled and the Pod's QoS was Guaranteed. |
2019-09-10 | v1.14.3-tke.3 | |
2019-09-06 | v1.14.3-tke.2 | Fixed the cve-2019-9512 & cve-2019-9514 HTTP/2 DDoS security vulnerabilities. Merged PR72914 to fix the issue where immediately creating and scheduling a Pod to the same node after its deletion could cause volume mount failures. Fixed the issue where creating containers on CentOS could cause cgroup leaks. |
Date | Version | Update |
2023-03-03 | v1.12.4-tke.32 | Fixed CVE-2022-3172. (kube-apiserver) Migrated 16+ webhook metrics and added the calling_webhook_timeout_error metric. (kube-apiserver)Migrated the etcd_request_duration_seconds metric. (kube-apiserver)Optimized the script for the self-adaptive node iptables mode to resolve misjudgment issues under certain conditions. (kube-proxy) Merged PR111635 to fix the issue where ipvs real server weights become zero after kube-proxy restarts. (kube-proxy) |
2022-04-13 | v1.12.4-tke.31 | Merged PR78428 to avoid writing warning messages when iptables rules are exported, which could cause kube-proxy to panic during recovery. (kube-proxy) |
2022-01-20 | v1.12.4-tke.30 | Supported binding the LB address to the ipvs network interface card. (kube-proxy) |
2021-05-06 | v1.12.4-tke.28 | Updated the startup method for kube-proxy when running in image mode to automatically adapt to the iptables mode of the node, supporting operating systems that run iptables in nf\\_tables mode by default. |
2020-12-28 | v1.12.4-tke.27 | Added metrics for QcloudCbs. (kube-controller-manager) Fixed the issue of extra spaces when the serial value is viewed while a cbs disk is mounted. (kubelet) |
2020-12-15 | v1.12.4-tke.26 | QcloudCbs supports BulkVolumeVerification. (kube-controller-manager) |
2020-11-17 | v1.12.4-tke.25 | Merged PR79495, fixing the issue where webhook invocation fails when a CRD has multiple versions. (kube-apiserver) |
2020-10-13 | v1.12.4-tke.24 | Merged PR93403, removing the erroneous log messages printed when kubelet updates Pod Conditions that do not belong to it. (kubelet) |
2020-08-04 | v1.12.4-tke.23 | Merged PR78881, fixing the bug where a default value was incorrectly added to pod.spec.container.SecurityContext.ProcMount. |
2020-07-28 | v1.12.4-tke.22 | Merged PR91277 to prevent the issue where CLB health checks cause kube-apiserver to generate a large number of TLS handshake error logs. (kube-apiserver) Merged PR91500 to fix the issue where the KUBERNETES_SERVICE_HOST environment variable is missing. (kubelet) |
2020-06-10 | v1.12.4-tke.21 | Merged PR73915 to prevent watchers from receiving events that occurred before they started watching. Merged PR91252 to ignore Pod Condition updates generated by other components, thereby avoiding unnecessary scheduling. Merged PR73915 to clean up error logs in kube-controller-manager, thereby preventing the Half-Blind SSRF attack associated with CVE-2020-8555. |
2020-06-04 | v1.12.4-tke.20 | Merged PR90260 to fix the missing network monitoring issue for containerd clusters. Merged PR79451 to fix the issue where kubelet does not recreate a SandBox after a creation failure when restartPolicy is Never. |
2020-05-18 | v1.12.4-tke.19 | Merged PR77802 to Disable graceful termination for UDP traffic. Merged PR68741 to fix the issue where, when a symlink for /var/lib/kubelet or a subpath is used, deleting a pod causes the host to fail to unmount volumes, resulting in mount point leaks and the pod remaining in a terminating state. tke can detect the maximum number of qcloudcbs volumes that can be mounted on a single node, but it does not support dynamically obtaining the maximum value. |
2020-04-14 | v1.12.4-tke.18 | Merged PR73401, PR73606, and PR76060 to delete DaemonSet Pods that were scheduled to non-existent nodes. Merged PR68619 to resolve the dirty data issue in the cpumanager. Merged PR87669 to fix CVE-2020-8552: the apiserver DoS attack issue. tke supports detecting the maximum number of qcloudcbs volumes that can be mounted on a single node (for version 1.12 and later, this limit is maxAttachCount-2, while for version 1.10, it defaults to 18). |
2020-02-14 | v1.12.4-tke.17 | The cbs V2 API has been upgraded to V3. cbs intree resolves the issue where unmounting disks continues when they do not exist, which generates a large number of invalid requests. |
2020-01-13 | v1.12.4-tke.16 | Merged PR2359 to fix the monitoring gap issue caused by the inability to obtain the docker root directory. Merged PR86583 to increase the log level when iptables does not support random-fully, thereby preventing excessive log generation. kube-scheduler supports dynamic log level configuration. Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for cbs is missing, enabling users to use cbs normally. Merged PR86230 to skip updating the scheduling of assumed pods during the pod scheduling process. |
2019-12-23 | v1.12.4-tke.15 | Reverted PR79036 to resolve the issue where cpu quota was disabled when cpumanager was enabled and the Pod's QoS was Guaranteed. |
2019-12-17 | v1.12.4-tke.14 | Added caching and timeout to metadata. Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. Prevent Pods from becoming not ready due to readiness probe failures during kubelet restarts. |
2019-11-28 | v1.12.4-tke.13 | cloud-provider added support for using the node name as the hostname. |
2019-11-18 | v1.12.4-tke.12 | Merged PR75622 to resolve the issue where changes to StatefulSets (sts) experience high latency (approximately 20 seconds) when the changes are synchronized to pods in clusters with a large number of sts workloads (>2000). |
2019-10-23 | v1.12.4-tke.11 | Merged PR79036 to fix the issue where CPU quota is disabled when the CPU Manager is enabled and a pod's QoS is Guaranteed. Merged PR72866 to add the --metrics-port command-line argument for kube-proxy and fix the bug where --metrics-bind-address cannot include a port. |
2019-09-06 | v1.12.4-tke.10 | Fixed the cve-2019-9512 & cve-2019-9514 HTTP/2 DDoS security vulnerabilities. Merged PR72914 to fix the issue where volume mount failures could occur when a Pod was immediately recreated and scheduled to the same node after deletion. Merged PR71834 to fix the issue where traffic could be directed to failed RS (Real Servers) when sessionAffinity is set to ClientIP in IPVS mode. |
2019-08-09 | v1.12.4-tke.9 | Fixed the issue where creating containers on CentOS could cause cgroup leaks. |
2019-08-08 | v1.12.4-tke.8 | Merged PR72118 to resolve the issue where a cbs-based StatefulSet fails to mount when rescheduled to the same Node. |
2019-07-17 | v1.12.4-tke.7 | |
2019-07-16 | v1.12.4-tke.6 | Resolved the compatibility issue between the tlinux kernel version and IPVS, and fixed the issue where CLB health checks fail in IPVS mode. |
2019-07-09 | v1.12.4-tke.5 | |
2019-06-25 | v1.12.4-tke.4 | Resolved the compatibility issue between the tlinux kernel version and IPVS. |
2019-06-17 | v1.12.4-tke.3 | |
2019-06-04 | v1.12.4-tke.2 | Merged PR74755 to fix the issue where the kubelet hangs. Merged PR69047 to fix the backward compatibility issue with node.Spec.Unschedulable. |
Date | Version | Update |
2021-05-06 | v1.10.5-tke.20 | Updated the startup method for kube-proxy when it is running in image mode to automatically adapt to the iptables mode of the node, supporting operating systems that run iptables in nf_tables mode by default. |
2020-06-10 | v1.10.5-tke.19 | Merged PR90260 to fix the missing network monitoring issue for containerd clusters. Merged PR91252 to ignore Pod Condition updates generated by other components, thereby avoiding unnecessary scheduling. Merged PR89794 to clean up error logs in kube-controller-manager, thereby preventing the Half-Blind SSRF attack associated with CVE-2020-8555. |
2020-05-18 | v1.12.4-tke.19 | Merged PR61549, adding volumeSpec data to the mountedPods cache to resolve the issue where pods could not be deleted normally when multiple pods used the same volume. |
2020-04-29 | v1.10.5-tke.17 | Merged PR75622 to resolve the issue where StatefulSet (sts) changes experience high latency (approximately 20 seconds) when they are synchronized to pods in clusters with a large number of sts workloads (>2000). |
2020-04-14 | v1.10.5-tke.16 | Merged PR68619 to resolve the dirty data issue in the cpumanager. Merged PR87669 to fix CVE-2020-8552: the apiserver DoS attack issue. tke supports detecting the maximum number of qcloudcbs volumes that can be mounted on a single node (for version 1.12 and later, this limit is maxAttachCount-2, while for version 1.10, it defaults to 18). |
2020-02-14 | v1.10.5-tke.15 | The cbs V2 API has been upgraded to V3. cbs intree resolves the issue where disks continue to be unmounted when they do not exist, which generates a large number of invalid requests. |
2020-01-13 | v1.10.5-tke.14 | Merged PR2359 to fix the monitoring gap issue caused by the inability to obtain the docker root directory. Merged PR86583 to increase the log level when iptables does not support random-fully, thereby preventing excessive log generation. kube-scheduler supports dynamic log level configuration. Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for cbs is missing, enabling users to use cbs normally. Merged PR86230 to skip updating the scheduling of assumed pods during the pod scheduling process. |
2019-12-23 | v1.10.5-tke.13 | Reverted PR79036 to resolve the issue where cpu quota was disabled when cpumanager was enabled and the Pod's QoS was Guaranteed. |
2019-12-13 | v1.10.5-tke.12 | kubelet does not delete a node when checking the externalID. Added caching and timeout to metadata. Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. Prevent Pods from becoming not ready due to readiness probe failures during kubelet restarts. |
2019-11-18 | v1.10.5-tke.11 | Removed the reverse probe for kube-controller-manager. |
2019-10-23 | v1.10.5-tke.10 | Merged PR79036 to fix the issue where CPU quota is disabled when the CPU Manager is enabled and a pod's QoS is Guaranteed. Merged PR72868 to add the --metrics-port command-line argument for kube-proxy and fix the bug where --metrics-bind-address cannot include a port. |
2019-09-06 | v1.10.5-tke.9 | Fixed the cve-2019-9512 & cve-2019-9514 HTTP/2 DDoS security vulnerabilities. Merged PR72914 to fix the issue where volume mount failures could occur when a Pod was immediately recreated and scheduled to the same node after deletion. Merged 67430 to fix the rollback of data structures when updateContainerCPUSet fails. |
2019-08-08 | v1.10.5-tke.8 | Merged PR72118 to resolve the issue where kubelet reports "resource name may not be empty" when mounting the same device immediately after unmounting. |
2019-07-17 | v1.10.5-tke.7 | |
2019-06-25 | v1.10.5-tke.6 | Resolved the compatibility issue between the tlinux kernel version and IPVS. |
2019-06-17 | v1.10.5-tke.5 | |
2019-03-19 | v1.10.5-tke.4 | |
2019-02-19 | v1.10.5-tke.3 | |
2018-09-28 | v1.10.5-tke.2 | Moved the logic for creating CLBs out of the controller-manager (implemented via an independent service controller). |
2018-09-27 | v1.10.5-tke.1 | Backported PR63321 to resolve the issue where pods with multiple business containers experience excessively long Terminating times. |
2018-09-21 | v1.10.5-qcloud-rev1 | When kubelet status update times out, the controller-manager probes the kubelet port. |
Date | Version | Update |
2020-01-13 | v1.8.13-tke.7 | Merged PR2359 to fix the monitoring gap issue caused by the inability to obtain the docker root directory. Worked around the issue where the device path (/dev/disk/by-id/virtio-xxx/...) for cbs is missing, enabling users to use cbs normally. |
2019-12-13 | v1.8.13-tke.6 | kubelet does not delete a node when checking the externalID. Added caching and timeout to metadata. Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. Prevent Pods from becoming not ready due to readiness probe failures during kubelet restarts. |
2019-11-18 | v1.8.13-tke.5 | Removed the reverse probing of kube-controller-manager. Add metrics for cbs pvc. |
2018-09-28 | v1.8.13-tke.2 | Moved the logic for creating CLBs out of the controller-manager (implemented via an independent service controller). |
2018-09-27 | v1.8.13-tke.1 | Disable kmem statistics to prevent cgroup quantity leakage. Reduce resourcequota conflicts triggered during pod creation. |
2018-09-21 | v1.8.13-qcloud-rev1 | When kubelet status update times out, the controller-manager probes the kubelet port. |
Date | Version | Update |
2019-12-17 | v1.7.8-tke.4 | kubelet does not delete a node when checking the externalID. Added caching and timeout to metadata. Fixed the issue where Pods exited due to the lxcfs upgrade on Ubuntu 16. Prevent Pods from becoming not ready due to readiness probe failures during kubelet restarts. |
2018-09-28 | v1.7.8-tke.2 | Resolved the conflict issue between the controller-manager and the external service controller. |
2018-09-27 | v1.7.8-tke.1 | Moved the logic for creating CLBs out of the controller-manager (implemented via an independent service controller). |
2018-09-21 | v1.7.8-qcloud-rev1 | When kubelet status update times out, the controller-manager probes the kubelet port. |
Apakah halaman ini membantu?
Anda juga dapat Menghubungi Penjualan atau Mengirimkan Tiket untuk meminta bantuan.
masukan