You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If you are deploying to kubernetes and have existing cluster with existing node pools, you might want to
A. Use pre-existing node_pool_x and assign all featureform pods to that with nodeSelector: label=node_pool_label
OR
B. Create new exclusive node_pool_featureform and assign all featureform pods to that pool rather than them getting randomly assigned to pre-existing node pools along with the new one.
Current Behavior
It is possible to set value for helm for nodeSelector but it isnt utilised for all services
If I'm lacking on kube/helm knowledge and there is currently a way to set it explicitly, please advice me.
Desired Behavior
All individual services' charts utilise the value nodeSelector that is set for helm deployment
Benefits
You might not have the option to create a new kube cluster but also have existing node pools that are configured elsewhere, and you wouldn't want to use in you featureform deployment. Being able to explicitly assign pods to a node pool would help.
Possible Implementation
Add
{{- with .Values.nodeSelector }}
nodeSelector:
{{- toYaml . | nindent 8 }}
{{- end }}
Also, if it isn't obvious, in a corporate setting you can't use the featureform/terraform configurations but you have to adapt it bc you already have a VPC, kube cluster, DNS etc.
The text was updated successfully, but these errors were encountered:
Description
If you are deploying to kubernetes and have existing cluster with existing node pools, you might want to
OR
Current Behavior
It is possible to set value for helm for nodeSelector but it isnt utilised for all services
https://github.com/search?q=repo%3Afeatureform%2Ffeatureform%20nodeselector&type=code
If I'm lacking on kube/helm knowledge and there is currently a way to set it explicitly, please advice me.
Desired Behavior
All individual services' charts utilise the value nodeSelector that is set for helm deployment
Benefits
You might not have the option to create a new kube cluster but also have existing node pools that are configured elsewhere, and you wouldn't want to use in you featureform deployment. Being able to explicitly assign pods to a node pool would help.
Possible Implementation
Add
to all services' yamls
then in terraform of you deployment
Additional Context
Also, if it isn't obvious, in a corporate setting you can't use the featureform/terraform configurations but you have to adapt it bc you already have a VPC, kube cluster, DNS etc.
The text was updated successfully, but these errors were encountered: