gpu-k8s: separate kubernetes version

This commit is contained in:
shenlan 2025-06-25 21:42:15 +08:00
parent 20b6647639
commit 35fca24f2e
3 changed files with 6 additions and 3 deletions

View File

@ -12,14 +12,16 @@ The role performs four main tasks:
4. **Verify GPU access** by deploying the official NVIDIA device plugin and running a small CUDA workload.
When `sealos_version` is set to `latest` (the default), the role automatically
fetches the most recent stable release from GitHub.
fetches the most recent stable release from GitHub. The Kubernetes image tag is
controlled separately via `kubernetes_version`, which defaults to `v1.25.16` but
can be overridden to any compatible release.
The following command is used to create the cluster (example with one master and one worker):
```bash
sealos run \
registry.cn-shanghai.aliyuncs.com/labring/kubernetes:<latest> \
registry.cn-shanghai.aliyuncs.com/labring/kubernetes:<kubernetes_version> \
registry.cn-shanghai.aliyuncs.com/labring/cilium:<cilium_version> \
registry.cn-shanghai.aliyuncs.com/labring/helm:<helm_version> \
--masters 172.16.11.120 \

View File

@ -1,5 +1,6 @@
# Default variables for gpu-k8s role
sealos_version: latest
kubernetes_version: v1.25.16
cilium_version: v1.13.4
helm_version: v3.9.4
master_ips: [] # List of up to three master node IPs

View File

@ -42,7 +42,7 @@
- name: Run sealos to create Kubernetes cluster
shell: |
sealos run \
registry.cn-shanghai.aliyuncs.com/labring/kubernetes:{{ sealos_version }} \
registry.cn-shanghai.aliyuncs.com/labring/kubernetes:{{ kubernetes_version }} \
registry.cn-shanghai.aliyuncs.com/labring/cilium:{{ cilium_version }} \
registry.cn-shanghai.aliyuncs.com/labring/helm:{{ helm_version }} \
--masters {{ master_ips | join(',') }} \