2023-01-05 |
v1.18.4-tke.34 |
DaemonSet pods can be run on super nodes when necessary. (kube-controller-manager)
Incorporates PR102576, which stops using RemoveAll when a remaining Pod volume is deleted to avoid deleting data related to the volume. (kubelet)
Allows users to reduce resource limits during in-place update. (kube-apiserver)
Fixes the issue where terminated containers are not deleted during log rotation. (kubelet)
Records the following metrics when list/watch too old, too large, and too many objects occur (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 |
Optimizes the scheduling of the number of local replicas on super nodes. |
2022-10-13 |
v1.18.4-tke.30 |
Incorporates PR112299, which optimizes the apiserver Gzip compression level. (kube-apiserver)
Fixes CVE-2022-3172 (kube-apiserver).
Adds the `calling_webhook_timeout_error` metric (kube-apiserver).
Ignores the error in case of `TKEDefaultQuota` conflicts and increases the number of workers to fix the `resource quota evaluates timeout` error (kube-apiserver).
Adds the `StatusDetaching` status for CBS disks (kube-controller-manager).
Supports DaemonSet for Serverless clusters (kube-apiserver). |
2022-09-07 |
v1.18.4-tke.28 |
Optimizes the list performance for large clusters (kube-apiserver).
Optimizes the scheduling of super nodes. (kube-scheduler)
Allows kube-system pods to be scheduled to prepaid super nodes. (kube-scheduler)
Optimizes super node HPA (kube-controller-manager).
Supports in-place update of Pod resources (kube-apiserver, kubelet).
Merges PR97348, which fixes the issue where the number for scaling is incorrect when `StabilizationWindowSeconds` is set for HPA (kube-controller-manager).
Merges PR108831, which fixes the issue where creating multiple Pods at a time leads to kubelet panic (kubelet). |
2022-07-27 |
v1.18.4-tke.26 |
CBS supports native nodes. (kubelet) |
2022-03-18 | v1.18.4-tke.23 | Merges pr92878, which allows to print alarm information only when setting the ownership for ConfigMap/Secret volumes takes more than 30 seconds, avoiding excessive log information. (kubelet) Merges pr106906, which detects whether the network storage volume subpath has been deleted, preventing the Pod from being in terminating status all the time. (kubelet) The TKE Serverless super nodes are ignored when the anti-affinity scheduling is performed based on the hostname. (kube-scheduler) Merges pr93026, which fixed the problem where DefaultPodTopologySpread cannot obtain replicaset information. (kube-scheduler) |
2022-01-20 | v1.18.4-tke.20 | TKE Serverless rescheduling optimization: Lower the score for super nodes that have been drained in the same availability zone. (kube-scheduler) The apiserver supports integration of ExternalName 556 type external services. (kube-apiserver) Supports binding the LB addresses to the ipvs ENIs. (kube-proxy) |
2021-12-09 | v1.18.4-tke.17 | Fixes the issue where kube-controller-manager's access to apiserver is restricted when there are a large number of volume attachment objects. (kube-controller-manager)Merges PR95650, so that HPA ignores deleted Pods when counting replicas. (kube-controller-manager)Fixes the inconsistency between TKE Serverless and frontend when calculating CPU resources. (kube-scheduler) |
2021-12-02 | v1.18.4-tke.16 | Fixes the bug when scheduling to super nodes. (kube-scheduler)Optimizes the super node scheduling algorithm. (kube-scheduler) |
2021-11-26 | v1.18.4-tke.15 |
- Merges pr96444, so that if an error occurs during RBAC policy sync, the operation will be retried. (kube-apiserver)
- Supports customized installation of other CNIs for added external hybrid cloud nodes. (kube-controller-manager)
- Supports binding cores by group for Android containers in cloud games. (kubelet)
- Supports extended scheduler Prebind and Unreserve operations. (kube-scheduler)
- Merges pr99336 to improve the node information sync mechanism upon kubelet startup. (kubelet)
- Fixes CVE-2021-25741 to block unauthorized access to server files over soft links. (kubelet)
- Optimizes the error message when scheduling times out due to CBS disk creation failure. (kube-scheduler)
- Optimizes gRPC logs to avoid printing too many logs when kubelet collects volume status. (kubelet)
- Avoids scheduling Pods using CBS to external CHC nodes. (kube-scheduler)
|
2021-08-23 | v1.18.4-tke.14 |
- When the TKE cluster adds nodes, it supports static IP. (kube-scheduler)
- When the TKE cluster adds nodes, if TKE Serverless static IP is matched, other pre-selected policies are skipped. (kube-scheduler)
- When the TKE cluster adds nodes, TKE Serverless node resource awareness rescheduling is optimized for TKE Serverless node scheduling, and TKE Serverless node priority model scheduling and preference/pre-selection policy for TKE Serverless nodes are optimized. (kube-scheduler)
- Records loaded IPVS kernel module to avoid kube-proxy crashes in IPVS mode. (kube-proxy)
- Avoids panic when an error occurs at the time of writing into cpu manager status file. (kubelet)
|
2021-07-22 | v1.18.4-tke.13 | Merges PR91859, which fixes the issue of kube-apiserver panic when the CRD type has only one letter. (kube-apiserver) |
2021-07-13 | v1.18.4-tke.12 | - When the TKE cluster adds nodes, it can perceive the remaining IPs in the subnet and schedule the right number of Pods to the super node at the time of batch scheduling of the Pods. (kube-scheduler)
- Supports collection of disk usage metrics of Containerd runtime. (kubelet)
- You can specify the Pod at the time of scaling 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 |
- Ports pr93370 to support CronJobControllerV2. (kube-controller-manager)
- Merges pr100376 to enable HTTP/2 health check, which prevents the issue that the underlying layer connection is closed but can still be used incorrectly. (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, kube-proxy, kubectl)
- Merges pr100317, which fixes the issue where CVE-2021-25735 node updates might bypass the Validating Admission Webhook. (kube-apiserver)
- When TKE cluster adds nodes, ComputeResource, TKE Serverless ClusterIP, and HPA are supported. (kube-controller-manager, kube-scheduler)
|
2021-04-02 |
v1.18.4-tke.8 |
- Merges pr97752, which fixes the issue where NewReplicaSet is displayed as
<none> when describing deployment (kubectl).
- Merges pr93808, which fixes the issue where unnecessary information is returned when
kube-scheduler --version is executed. (kube-scheduler)
- Merges pr91590, which fixes the issue of warning that the port has been allocated when using the multiprotocol service of NodePort type (kube-apiserver).
- Merges pr98262, which allows you to use kube-controller-manager to dynamically adjust the log level. (kube-controller-manager)
- Merges pr95154, which fixes the issue where kube-scheduler snapshot contains the nodes being deleted. (kube-scheduler)
- Merges pr95711, which fixes the issue where kubectl drain command occupies too much CPU. (kubectl)
- Merges pr96602, which fixes the issue where apiserver memory leaks before or after the time gaps. (kube-apiserver)
- Merges pr97023, which deletes the related metadata directory when unmounting an emptyDir type volume (kubelet).
- Merges pr97527, which fixes the issue where map access operations are not synchronized in cpumanager (kubelet).
- Merges pr100190, which automatically deletes the volume directory left by orphaned Pod (kubelet).
- Merges pr92614, when all containers of the Pod whose restart policy is RestartPolicyOnFailure exit successfully, no new sandbox will be created (kubelet).
- Merges pr94833, which fixes the issue where the image tag does not match in status when Pod image has multiple tags (kubelet).
|
2020-12-28 |
v1.18.4-tke.6 (ARM clusters are supported starting from this version) |
- Adds metrics to QcloudCbs. (kube-controller-manager)
- Fixes the issue where extra space exists in the value of serial when mounting CBS disk. (Kubelet)
|
2020-12-21 |
v1.18.4-tke.5 |
- Merges pr94712, which fixes CVE-2020-8564 - fixes the issue when the file format is incorrect and logLevel >= 4, Docker configuration leaks. (kubelet)
- Merges pr95316, which fixes CVE-2020-8565 - fixes the issue where incomplete fix for CVE-2019-11250 resulting in log token leak. (logLevel >= 9) (kube-apiserver, kubectl)
- Merges pr95245, which fixes CVE-2020-8566 - fixes the issue where Ceph RBD adminSecrets is exposed in the log when loglevel >= 4. (kube-controller-manager)
- Fixes the issue where restarting kubelet causes failure of Pod readiness check. (kubelet)
- Merges pr90825, which fixes the issue where the pop operation of the fifo queue in client-go might be stuck due to race condition, which causes the pod to remain in the pending state. (kubelet)
- The scheduler supports super nodes. (kube-scheduler)
- The kube-controller-manager supports super nodes. (kube-controller-manager)
- Sets the instance-type label based on the actual model of the node, instead of being fixed as QCLOUD. (kubelet)
- Adds the CBS to OpenAPI. (kube-apiserver)
- Merges pr91126, which fixes the issue where the scheduler cache is inconsistent when Pod has the same name but different UID. (kube-scheduler)
- Merges pr93387, which fixes the issue where the daemonset pod can not be scheduled to nodes due to the disorder of node cache information in the scheduler. (kube-scheduler)
- Merges pr89465, which fixes the issue where the 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 |
- Merges pr89629, which fixes the issue where the container that mounts the subpath would fail to restart after the configmap is changed. (kubelet)
- QcloudCbs supports BulkVolumeVerification. (kube-controller-manager)
- Merges pr94430, which fixes the issue where the client-go reflector could not detect the "Too large resource version" error (kubelet).
|
2020-08-12 |
v1.18.4-tke.2 |
- Merges pr93403, which removes the printed error information of pod condition irrelevant to the kubelet during kubelet update. (kubelet)
|
2020-08-04 |
v1.18.4-tke.1 |
- Reverts pr63066, which fixes the issues of LB health check and IPVS. (kube-proxy)
- Merges pr72914, which fixes the issue where mounting might fail if you delete a pod, create a new one, and schedule it to the same node. (kube-controller-manager)
- Fixes the issue where creating containers in CentOS results in cgroup leakage. (kubelet)
- Fixes the issue where upgrading lxcfs in Ubuntu 16 causes pods to exit. (kubelet)
- metadata adds cache and timeout. cloud-provider now supports using node name as hostname. (kubelet)
- metadata adds local cache. (kubelet)
- Incorporates CBS and relevant fixing code. (kubelet)
- Merges pr90260, which fixes the issue of missing monitoring records for containerd cluster networks. (kubelet)
- TKE can perceive the maximum number of qcloudcbs that can be mounted to a single node. In 1.12 and later versions, the value is maxAttachCount-2. In version 1.10, the value is 18 by default. (kube-scheduler)
- Fixes the issue where CBS intree continues to unmount a non-existent disk, causing numerous invalid requests. (kubelet)
- Merges pr2359, which fixes the issue with missing monitoring records when the system is unable to obtain docker root. (kubelet)
- kube-scheduler now supports dynamic logging level configuration. (kube-scheduler)
- Produces a workaround for the missing CBS device path (/dev/disk/by-id/virtio-xxx/...) issue that prevents some users from accessing CBS properly. (kubelet)
- TKE can perceive the maximum number of qcloudcbs that can be mounted to a single node. The kubelet side will not patch node. (kubelet)
- Merges pr89296, so that the log will not record whether the iptables random-fully parameter is enabled. (kube-proxy)
- Fixes the aws issue, pr92162. (kubelet)
- Merges pr91277, which prevents the issue of large numbers of TLS handshake error logs generated by kube-apiserver as a result of CLB health checks. (kube-apiserver)
- Merges pr91500, which fixes the issue of missing environmental variables of KUBERNETES_SERVICE_HOST. (kubelet)
- Merges 92537, which fixes the issue where client-go reflector could not recover from the error "Too large resource version". (kube-apiserver, kube-controller-manager, kube-scheduler, kubelet, and kube-proxy)
- Merges pr92969, which fixes the issue where CVE-2020-8559 privilege escalation from an invaded node results in invasion into other nodes. (kube-apiserver)
- Merges pr92921, which fixes the DOS attack issue where CVE-2020-8557 exhausts the disk space by writing into “/etc/hosts”. (kubelet)
|
Was this page helpful?