To be compliant with normal standards in the industry we should not have pods without security context or resources limits.
This issue block us form using Ollama operator in production workflows.
Two proposals for the statefulset:
1 - Apply same security context definition and same resources limits from the models pods
Cons:
- We cannot fine tune resources limits based in the usage for each pod
Pros:
- Security context is normally the same for every pod
2 - Define statefulset security context definition and resources limited (for instance: extraSecurityContext, extraContainerSecurityContext and extraResources fields in model manifest)
Cons:
- Might imply larger code change
Pros:
- Allow deeper customization
To be compliant with normal standards in the industry we should not have pods without security context or resources limits.
This issue block us form using Ollama operator in production workflows.
Two proposals for the statefulset:
1 - Apply same security context definition and same resources limits from the models pods
Cons:
Pros:
2 - Define statefulset security context definition and resources limited (for instance:
extraSecurityContext,extraContainerSecurityContextandextraResourcesfields in model manifest)Cons:
Pros: