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
"the kubeai controller pod is supposed to pick up changes to configs without restarting right (e.g. adding a new resource profile via the helm values file)? I've created a resource profile and new model crd but it doesn't seem to pick it up and spin up a new replica. Nothing of note in the logs, and I don't want to restart the controller pod since there's active usage"
Today the resource profiles are defined in a configmap. KubeAI however does not automatically pick up any changes in this configmap.
We should consider that modifying ResourceProfile is a common operation. So depending on pod restarts isn't ideal. A pod restart could cause issues to existing connections.
The text was updated successfully, but these errors were encountered:
samos123
changed the title
Modifying helm values resourceProfile should auto refresh kubeai controller
Modifying kubeai configmap should auto refresh kubeai controller
Aug 30, 2024
samos123
changed the title
Modifying kubeai configmap should auto refresh kubeai controller
Modify Resource Profiles without requiring pod restart
Sep 17, 2024
…ield (#214)
Restart KubeAI to pick up changes to configmap.
See:
https://helm.sh/docs/howto/charts_tips_and_tricks/#automatically-roll-deployments
Hot-reload was considered - however it gets complicated b/c of
long-running processes that get started such as messengers and
autoscaling. We can punt on that until later.
Related to #151 which can be implemented seperately.
Also added `.messaging` to ConfigMap (it was omitted).
Relevant quote:
"the kubeai controller pod is supposed to pick up changes to configs without restarting right (e.g. adding a new resource profile via the helm values file)? I've created a resource profile and new model crd but it doesn't seem to pick it up and spin up a new replica. Nothing of note in the logs, and I don't want to restart the controller pod since there's active usage"
Today the resource profiles are defined in a configmap. KubeAI however does not automatically pick up any changes in this configmap.
We should consider that modifying ResourceProfile is a common operation. So depending on pod restarts isn't ideal. A pod restart could cause issues to existing connections.
The text was updated successfully, but these errors were encountered: