-
-
Notifications
You must be signed in to change notification settings - Fork 38
Description
Hi all -- thanks for bringing us this project, it's just great!
I am using v0.6.1 I have updated an existing PVC with the following annotation (+ shows the addition):
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: foobar
annotations:
gcs.csi.ofek.dev/project-id: {{ .Values.gcloud.project }}
gcs.csi.ofek.dev/location: {{ .Values.gcloud.region }}
gcs.csi.ofek.dev/bucket: {{ .Values.code }}-foobar
gcs.csi.ofek.dev/uid: "0"
gcs.csi.ofek.dev/gid: "0"
+ gcs.csi.ofek.dev/limit-ops-per-sec: "-1"
The PVC has been correctly updated, and I killed the Pod mounting this PVC so that it has been re-mounted.
Looking at the logs in the csi-gcs Pod I see the following when the volume is mounted:
"Mounting cmd (mount) with arguments ([-t gcsfuse -o key_file=/tmp/keys/122073239,allow_other,dir_mode=0775,file_mode=0664,uid=0,gid=0 us-foobar /var/lib/kubelet/pods/63d32ac5-1aac-4338-8d0d-0f7ac3d09cc9/volumes/kubernetes.io~csi/pvc-edb69964-0e18-4716-a8b4-92edf8bdee19/mount])"
I can see how some of the annotations that were originally in the PVC are showing up in the gcsfuse command (e.g. uid or gid) but I can't see the limit-ops-per-sec option anywhere.
So I wonder: are these kind of updates to the PVC actually picked up when re-mounting the volume, or do they apply only when the PVC is created the first time?
Thanks
Update
I have created a fresh new PVC and mounted it on a deployment and this csi-gcs is logging the following:
Mounting cmd (mount) with arguments ([-t gcsfuse -o
key_file=/tmp/keys/030309528,allow_other,dir_mode=0775,file_mode=0664,
uid=0,gid=0,limit_ops_per_sec=-1 foobar /var/lib/kubelet/pods/0130f46a-7ae9-4dd2-acab-d3a697f1f5ec/volumes/kubernetes.io~csi/pvc-d7a805a1-abd1-43a8-b473-cf1e06e005dc/mount])
it seems that creating a new PVC will use the right gcsfuse flags, but updating an existing PVC doesn't guarantee that gcsfuse flags are updated.