The planned upgrade for Tencent Kubernetes Engine (TKE) cluster components provides users with customizable cluster upgrade windows and upgrade sequences based on user maintenance windows, exclusion items, and cluster orchestration configurations. This document details the planned upgrade process and grayscale policies.
Upgrade Process
The planned upgrade of TKE cluster components consists of the following phases:
1. Generation of Candidate Clusters for Upgrade
During a component upgrade task, Tencent Cloud selects clusters that meet the following upgrade criteria and adds them to the cluster component upgrade queue:
The cluster is in normal status (not in deleted or arrears status).
Components are in normal operation.
Cluster and component configurations meet upgrade requirements.
The cluster has a configured maintenance window, and there are available maintenance time windows within a certain period.
2. Upgrade Task Planning
Tencent Cloud divides the clusters in the upgrade queue into multiple batches based on a specific grayscale sequence. In each batch, clusters are divided into two types based on whether they are in an enabled release sequence, and cluster component upgrade plans are scheduled based on different policies.
For clusters not in an enabled release sequence, Tencent Cloud obtains the latest available maintenance period by combining the cluster maintenance windows and exclusion items to generate the cluster component upgrade plan. For clusters in an enabled release sequence, Tencent Cloud refers to the user cluster release sequence, combined with the cluster maintenance windows and exclusion items, to generate the upgrade plan for each cluster component in the release sequence. Tencent Cloud follows the cluster release order and interval specified in the release sequence to ensure the upgrade sequence of cluster components, while planning the cluster component upgrade time period based on the maintenance windows and exclusion items.
3. Pre-upgrade Check
Before executing the component upgrade, Tencent Cloud performs a pre-upgrade check to ensure that the status and configuration of the current component meet upgrade requirements. Tencent Cloud configures a series of different pre-check items for different components, including operation status check, key configuration item check, key parameter check, simulated scheduling check, and abnormal event check.
4. Upgrade Task Scheduling and Execution
Tencent Cloud schedules and executes the cluster component upgrade task within the planned upgrade time period of the cluster. If the cluster cannot be scheduled for upgrade within the planned period, Tencent Cloud will skip the upgrade of the cluster component and mark the cluster upgrade as "failed". Users can cancel the upgrade task before the execution of the cluster component upgrade task.
5. Post-upgrade Check
After the successful execution of the cluster component upgrade, Tencent Cloud automatically performs a post-upgrade check to verify that the operation status and features of the components are normal. Tencent Cloud configures a series of different post-check items for different components, including operation status check, abnormal event check, and key metric check.
Upgrade Grayscale Policies
Upgrading by the Default Grayscale Policies
The Tencent Cloud planned upgrade uses the grayscale policies by default, performing grayscale batch upgrades for clusters by region. The grayscale sequence is as follows:
1. Hong Kong (China), Macao (China), and Taiwan (China), and regions outside the Chinese mainland, such as Hong Kong (China), Tokyo, and Singapore.
2. Regions in the Chinese mainland (Beijing, Shanghai, and Guangzhou excluded), such as Chengdu, Wuhan, and Hangzhou
3. Beijing region
4. Shanghai region
5. Guangzhou region
For clusters within the same region, Tencent Cloud performs grayscale batch upgrades based on cluster scale. For TKE clusters, clusters are divided into batches based on cluster specifications, and those with smaller specifications are upgraded first. For TKE Serverless clusters, clusters are divided into batches based on the number of Pods within each cluster, and those with fewer Pods are upgraded first.
If clusters under the user account are not in any enabled release sequence when Tencent Cloud plans the cluster component upgrade task, these clusters will be divided into batches based on the default grayscale policy to schedule the cluster component upgrade plan.
Upgrading by the Release Sequence
Users can classify clusters and define the release order of different types of clusters based on cluster tags (refer to Cluster Orchestration). Tencent Cloud matches clusters to their release sequences based on cluster tags and determines the list of clusters in each release batch and the minimum interval between different release batches based on the cluster tags specified for each release phase in the release sequence.
Upgrade Frequency and Plan
Tencent Cloud pushes planned upgrade tasks for cluster components based on the version distribution of user clusters and components, including version upgrades for earlier version components and fixes for component issues. Before generating a planned upgrade task for cluster components, Tencent Cloud automatically checks whether user clusters meet the upgrade requirements and creates the planned upgrade task at least three days in advance. Users can view created upgrade tasks in the last three months under their account on the Ops Tasks page in the TKE console. Upgrade Notice
Users can view the upgrade tasks in the last three months under their account on the Ops Tasks page in the TKE console, and check the clusters involved in the upgrade task as well as the upgrade results through task details. Upgrade Tasks
Users can view component upgrade tasks in the last three months under their account on the Upgrade Tasks page, including the following task information:
Ops task: name of the upgrade task
Component: name of the component to be upgraded
Associated resources: cluster resources associated with the component upgrade
Planned upgrade time: the planned start time of the cluster upgrade task
Upgrade impact: possible impacts of component upgrade, including new feature release and issue fixes
Creation time: creation time of the upgrade task
Upgrade Plan Details
Users can view the clusters associated with the upgrade task and the upgrade status on the Upgrade Plan Details page, including the following information:
Cluster ID: ID of the cluster to be upgraded
Cluster name: name of the cluster to be upgraded
Estimated upgrade time: estimated start time of the cluster component upgrade
Start time: start time of the cluster component upgrade
End time: completion time of the cluster component upgrade
Execution status: status of the cluster component upgrade
Reason: reason for upgrade failure
Operation Steps
Managing Ops Tasks
1. Log in to the TKE console and go to the Plan Upgrade page. 2. Go to the Ops task subpage to view the upgrade tasks created in the last three months under the account. 3. Click Details to view the cluster upgrade details in the upgrade tasks.
4. Click Cancel on the right of the upgrade plan details to cancel the upgrade of the cluster component in the upgrade task. After cancellation, the cluster component upgrade will be ignored in the current upgrade task.