Due to the nature of EBS volumes they must be created in the same AZ as the EC2 instance that uses them, and this doesn’t seem to be handled by Rancher at all.
If I create a workload Rancher will create the relevant pods on one of my workers. If I configure a persistent volume for said workload, the EBS volume will be created in a seemingly random AZ with no guarantees that it’s actually accessible by the instance running my pod.
Is there a reasonable way of getting this to work? At the moment we’ve had to resort to specifying scheduling rules so that we know exactly which AZ the pods will be created in, and to create multiple storage classes (one for each AZ) from which we choose the correct one.