升级指南

LLMOS 包含一个内置的升级检查机制，每 60 分钟自动检查新版本。升级检查的设置可以在全局设置页面上进行配置。

管理员用户可以查看所有可用版本，并通过命令行手动升级到新版本。

信息

LLMOS不支持降级。

通过命令行升级

登录到其中一个管理节点，或设置 KUBECONFIG 环境变量指向正确的 LLMOS 集群。
```
export KUBECONFIG=/path/to/llmos-kubeconfig
```

运行以下命令将集群升级到较新版本：

kubectl apply -f - <<EOF
apiVersion: management.llmos.ai/v1
kind: Upgrade
metadata:
  name: upgrade-v011-rc1
spec:
  version: v0.1.1-rc1 # 要升级到的版本
  registry: "docker.io/llmosai" # 如有需要，覆盖默认注册表（例如，私有注册表）
EOF

等待升级完成。成功的升级 YAML 看起来如下（status.state 为 Complete）：

apiVersion: management.llmos.ai/v1
kind: Upgrade
metadata:
  annotations:
    kubectl.kubernetes.io/last-applied-configuration: |
      {"apiVersion":"management.llmos.ai/v1","kind":"Upgrade","metadata":{"annotations":{},"name":"llmos-upgrade-test"},"spec":{"registry":"docker.io/llmos-dev","version":"v0.1.1-rc1"}}
  creationTimestamp: "2024-09-23T02:32:24Z"
  generation: 1
  labels:
    llmos.ai/latest-upgrade: "true"
  name: llmos-upgrade-test
  resourceVersion: "9348507"
  uid: ca703060-73a0-4b32-8b93-9266c8c7a28f
spec:
  version: v0.1.1-rc1
status:
  PreviousKubernetesVersion: v1.31.0+k3s1
  appliedVersion: v0.1.1-rc1
  completeTime: "2024-09-23T02:36:36Z"
  conditions:
  - lastUpdateTime: "2024-09-23T02:36:36Z"
    message: Upgrade completed
    reason: Ready
    status: "True"
    type: Completed
  - lastUpdateTime: "2024-09-23T02:32:24Z"
    message: Chart repo is ready for upgrade v0.1.1-rc1
    reason: Ready
    status: "True"
    type: ChartsRepoReady
  - lastUpdateTime: "2024-09-23T02:36:35Z"
    message: All managed system addons are ready
    reason: Ready
    status: "True"
    type: ManagedAddonsIsReady
  - lastUpdateTime: "2024-09-23T02:35:40Z"
    message: Manifest upgrade is ready
    reason: Ready
    status: "True"
    type: ManifestUpgradeComplete
  - lastUpdateTime: "2024-09-23T02:36:36Z"
    message: All nodes upgraded to version v0.1.1-rc1(v1.31.1+k3s1)
    reason: Ready
    status: "True"
    type: NodesUpgraded
  managedAddonStatus:
  - complete: true
    disabled: false
    jobName: helm-install-gpu-operator
    name: gpu-operator
  - complete: true
    disabled: false
    jobName: helm-install-kuberay-operator
    name: kuberay-operator
  - complete: true
    disabled: false
    jobName: helm-install-llmos-operator-redis
    name: llmos-operator-redis
  - complete: true
    disabled: false
    jobName: helm-install-rook-ceph
    name: rook-ceph
  planStatus:
  - complete: true
    lastUpdateTime: "2024-09-24T01:48:31Z"
    latestHash: 962f583a45a4301325b824255d3c2232b5372caed16741717e5317db
    latestVersion: v0.1.1-rc1
    name: llmos-upgrade-test-agent
  - complete: true
    lastUpdateTime: "2024-09-24T02:42:45Z"
    latestHash: 962f583a45a4301325b824255d3c2232b5372caed16741717e5317db
    latestVersion: v0.1.1-rc1
    name: llmos-upgrade-test-server
  previousVersion: v0.1.0
  startTime: "2024-09-23T02:32:24Z"
  state: Complete
  upgradeJobs:
  - complete: true
    helmChartName: llmos-crd
    lastUpdateTime: "2024-09-23T02:34:55Z"
    name: helm-install-llmos-crd
  - complete: true
    helmChartName: llmos-operator
    lastUpdateTime: "2024-09-23T02:35:43Z"
    name: helm-install-llmos-operator

添加新版本

新版本可以在upgrade-check-url中找到，如果您不想启用默认的升级检查器，可以通过手动创建新的 Version 资源来添加新版本，如下所示。

kubectl apply -f - <<EOF
apiVersion: management.llmos.ai/v1
kind: Version
metadata:
   name: v0.1.1-rc1
spec:
   minUpgradableVersion: v0.1.0 # 可升级到此版本的最低版本
   kubernetesVersion: v1.31.0+k3s1 # 新版本中包含的更新的 k8s 版本
   releaseDate: "2024-09-30"
   tags: ["preview"]
EOF

注意

使用此方法时，您需要负责创建正确的 Version 资源，且除非您完全了解，否则不要修改从升级检查 URL 复制的 spec 参数。

通过命令行升级​

添加新版本​

通过命令行升级

添加新版本