CrashLoopBackOff after draining node about csi-hetzner HOT 3 CLOSED

With further investigation it seems CSI on hetzner is not compatible with v1.13+ yet.
More information : digitalocean/csi-digitalocean#113
Any ETA for update ?

from csi-hetzner.

I'm unfortunately not able to reproduce the issue yet. I setup a test cluster on v0.12.3, installed the CSI Driver, upgrade to k8s v0.13.1 and played around with draining and undraining nodes: no issues with both the CSI controller as the nodes using PVCs.

Also setting up directly on a v0.13.1 cluster does not seem to give any issues.

I'm wondering, as zenjoy/csi-hetzner/driver/driver.go:91 (https://github.com/zenjoy/csi-hetzner/blob/master/driver/driver.go#L91) points to a part in the code related to a response from the Hetzner API: maybe there was a glitch in the API? Because it's very weird to get a nullpointer exception at that point in the code unless the Hetzner API returned no results when querying for the server details?

Btw, @thcyron from Hetzner confirmed they will release an officially supported CSI Driver after the holidays: apricote/hcloud-csi-driver#1 (comment) It might be a good idea at that point to migrate to using their driver as this was only meant as a simple port of the DigitalOcean driver to get started and do some testing during the Hetzner Volumes Beta.

from csi-hetzner.

That was a problem with Hetzner API as you said.
Thank you

from csi-hetzner.