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 hcloud-csi-controller deployment has no node anti-affinity for nodes labeled with instance.hetzner.cloud/is-root-server=true by default. This means that the pod can get deployed on robot servers, rendering it useless, as it cannot access the API.
The hcloud-csi-node DaemonSet already has anti-affinity for labeled robot nodes, it should probably be added to the deployment aswell.
The text was updated successfully, but these errors were encountered:
Its not very well documented, but the controller does not need to run on Cloud Nodes if you set the HCLOUD_VOLUME_DEFAULT_LOCATION environment variable to your preferred default location. By default the controller uses the location of the Cloud Server its running on (through the metadata service), but if you configure it you can run it anywhere you like.
This is something that we should make easily configurable through the Helm Chart and change the default to be more user friendly. I will update the issue title as such.
apricote
changed the title
hcloud-csi-controller can be deployed on robot nodes, breaking it.
fix(chart): Make default values work with cloud/dedicated hybrid clusters
Feb 9, 2024
The
hcloud-csi-controller
deployment has no node anti-affinity for nodes labeled withinstance.hetzner.cloud/is-root-server=true
by default. This means that the pod can get deployed on robot servers, rendering it useless, as it cannot access the API.The
hcloud-csi-node
DaemonSet already has anti-affinity for labeled robot nodes, it should probably be added to the deployment aswell.The text was updated successfully, but these errors were encountered: