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
We are making use of taints and tolerations to group different node-roles inside our cluster. This prevents brupop to schedule updates for any tainted nodes, as it is excluded from running an agent there.
Also, it would be nice to be able to control the scheduling of the central components (apiserver & controller) to specific nodes by exposing nodeSelector and pod(Anti)Affinity throughout the helm chart. This would enable us to group our "infra-relevant" deployments on a seperated group of nodes.
I would gladly create a PR if required, but it comes down to adding a few lines of default helm templating, like
# ...{{- with .Values.tolerations }}tolerations:
{{- toYaml . | nindent 8 }}{{- end }}# ...
for the agent daemonset, api-server and controller deployment. And something like
# ...{{- with .Values.nodeSelector }}nodeSelector:
{{- toYaml . | nindent 8 }}{{- end }}# ...
for api-server and controller deployment.
The text was updated successfully, but these errors were encountered:
@hofalk we're planning to release this in Brupop 1.3.0, which you can track in #508.
I was curious if you wouldn't mind weighing in on the implementation (#513). I think your suggested use-case of separating infrastructure pods onto their own hardware is probably the most likely, but I added individualized controls for each component to allow for other bespoke setups.
I probably should have asked before merging 😅, but happy to circle back if you have any suggestions or concerns.
Image I'm using: 1.2.0
Issue or Feature Request: Feature Request
We are making use of taints and tolerations to group different node-roles inside our cluster. This prevents brupop to schedule updates for any tainted nodes, as it is excluded from running an agent there.
Also, it would be nice to be able to control the scheduling of the central components (apiserver & controller) to specific nodes by exposing
nodeSelector
andpod(Anti)Affinity
throughout the helm chart. This would enable us to group our "infra-relevant" deployments on a seperated group of nodes.I would gladly create a PR if required, but it comes down to adding a few lines of default helm templating, like
for the agent daemonset, api-server and controller deployment. And something like
for api-server and controller deployment.
The text was updated successfully, but these errors were encountered: