Closed macgregor closed 1 year ago
First of all, please accept my apologies for the very late response.
Well, I'm not sure anybody would really need iSCSI volumes on control-plane nodes because none of the predefined workloads (e.g. Core DNS) requires them in the first place, but... you are right that this should be supported anyway so that it could be used in similar use cases. Would you mind sending a PR?
I ended up switching to democratic-csi and simple nfs volumes due to synology's rather severe LUN limitations and some other usability issues i encountered trying to use iscsi.
In release 0.6.0, support for affinity, nodeSelector and tolerations was added.
First off, thanks a ton for creating this. Great job! Hopefully the synology folks pick it up and run with it.
Many helm charts provide the ability to set tolerations and node affinity of the components they create. For example, if I wanted to make sure the synology-csi controller runs on my control plane nodes and not worker nodes I'd add the following values (particular values are specific to my infrastructure, obviously):
This may even be considered a non-starter for any but the most trivial kubernetes clusters. I'm not sure if some of the synology-csi components are meant to be running on all nodes on a cluster. It seems to be working fine on my two node cluster even though everything is deployed to the worker node due to taints on the master, but im still getting things up and running.
Here is an example in oath2-proxy.