Dilatation NodeGroup
Introduce a Node Group Upgrade feature to allow the upgrade of kubelet and container runtime versions on nodes. For GPU nodes, this also includes upgrades to the nvidia-container-toolkit version. Before performing a dilatation operation, the K8S Cluster CCE will conduct pre-checks. Node groups must pass these checks before proceeding with dilatation.
Note
- During node group dilatation, only dilatation operations are supported, and scale-down operations are not allowed. Nodes will not be removed during this process.
- Node group dilatation is performed in-place by default. Certain scenarios may require node draining before dilatation, such as when transitioning the container runtime from Docker to Containerd.
- If your cluster contains orphaned nodes (worker nodes not managed by a node group), you can refer to Add External Existing Nodes to include them in a node group for dilatation.
- A single account can simultaneously dilate up to 10 node groups.
Upgrade instructions
Node group dilatation covers updates to kubelet, container runtime, and nvidia-container-toolkit versions.
- Kubelet dilatation: Allows dilatation of all nodes within a node group to align with the kubelet version displayed in the current cluster control panel.
- Container runtime dilatation: Enables all nodes in a node group to update to the latest container runtime version supported by the cluster, including transitions from Docker to Containerd.
- Nvidia-container-toolkit dilatation: Allows all nodes in a node group to update to the latest supported version of the nvidia-container-toolkit.
Operation steps
- Sign in to Cloud Container Engine (CCE) console.
- Select Cluster List in the left navigation bar.
- On the Cluster List page, click on the desired cluster to open the cluster management page.
- Select Node Management > Node Group from the left navigation bar in the cluster management page.
- In the target node group's operation column, select More > Dilatation to enter the node group dilatation page and complete the dilatation configuration.
| ConfigMap | Required/Optional | Configuration |
|---|---|---|
| Dilatation object | Required | Select the node object to be dilated. |
| Dilatation scope | Required | Select the scope of nodes to be dilated. All nodes: By default, all eligible nodes in a node group are selected for dilatation. Specified nodes: Allows selecting one or multiple nodes within a node group for dilatation, with the option to redilate after the initial operation. |
| Drain node | Optional | Decide whether to perform a drain operation to evict pods from the node before starting node dilatation. |
| Maximum node count per batch execution | Required | Node dilatation within a node group is carried out in batches. Set the maximum number of nodes to undergo parallel dilatation per batch (default: 1, maximum: 10). |
| Auto-suspend policy | Required | Configure the suspend policy during dilatation. Suspend after the first batch: The task pauses automatically after completing the first batch, requiring a manual trigger to continue. Suspend after each batch: The task pauses automatically after completing every batch, needing a manual trigger to proceed. Do not suspend: The task continues seamlessly between batches without interruption. |
| Interval between batches | Required | When Auto-Suspend Policy is set to No Suspend, the interval between batches must be set. Not Set: Proceed directly to the next batch. Set Interval: Dilatation will automatically proceed to the next batch after the interval elapses. |
| Interval | Required | When Batch Interval is set to Set Interval, the interval between successive batch dilatation during the dilatation process must be specified, defaulting to 5 min with a set range of 5-120 min |
- Click Pre-check. Node dilatation begins after all dilatation checks pass.
-
During dilatation, you can carry out the following operations.
- Suspend: Suspend the dilatation. Click Suspend to temporarily suspend the node group dilatation, while nodes undergoing dilatation will continue until completion.
- Continue: Proceed the dilatation. Click Continue to proceed dilating remaining nodes of the node group.
- Cancel: Cancel the dilatation, which requires suspending the dilatation first. Click Cancel to continue subsequent node dilatation.
- After dilatation is complete, click the Node Name on the Node page to verify whether the kubelet version, container runtime version, and other details on the Basic Information page meet expectations.
Note:
The suspended state represents an intermediate phase during node group dilatation. It is advisable to avoid performing operations on the cluster during this period and to finalize the dilatation process promptly. The cluster will automatically terminate the dilatation process for node groups in intermediate states after 7 days, cleaning up all dilatation-related events and log information.
