helm chart 用的是 Helm chart for Ollama on Kubernetes https://github.com/otwld/ollama-helm
镜像用的是 mazurkatarzyna/ollama-deepseek-r1-7b,其中内置 deepseek-r1:7b 模型,之前用 docker compose 成功部署过这个镜像,详见博文 https://www.cnblogs.com/dudu/p/18705441
在 k8s 集群上用 helm部署后 pod 成功启动,但进入容器内查看却没有任何模型
# kubectl exec -it deployment/ollama-deepseek -- ollama list
NAME ID SIZE MODIFIED
请问如何解决这个问题?
heml 清单文件如下
image:
repository: mazurkatarzyna/ollama-deepseek-r1-7b
tag: latest
podLabels:
app: ollama-deepseek
tolerations:
- key: "dedicated"
operator: "Equal"
value: "deepseek"
effect: "NoSchedule"
nodeSelector:
dedicated: deepseek
ollama-helm chart 默认会 mount /root/.ollama volume
volumeMounts:
- name: ollama-data
mountPath: {{ .Values.ollama.mountPath | default "/root/.ollama" }}
{{- if .Values.persistentVolume.subPath }}
subPath: {{ .Values.persistentVolume.subPath }}
{{- end }}
通过下面的清单 mount 一个不存在的路径
ollama:
mountPath: "/root/.empty"
问题就解决了
# kubectl exec -it deployment/ollama-deepseek -- ollama list
NAME ID SIZE MODIFIED
deepseek-r1:7b 0a8c26691023 4.7 GB 2 weeks ago