CSI base Object Storage driver/plug-in. Currently, the driver supports s3fs and rclone mounters.
For building the driver docker
and GO
should be installed on the system
-
Install latest Go
-
Set the
GOPATH
environment variable. -
Build the driver image
$ mkdir -p $GOPATH/src/github.com/IBM $ cd $GOPATH/src/github.com/IBM/ $ git clone https://github.com/IBM/ibm-object-csi-driver.git $ cd ibm-object-csi-driver
export RHSM_USER=<RHSM_USER> export RHSM_PASS=<RHSM_PASS> make buildimage
An image named ibm-object-csi-driver:latest
is created. Please retag and push the image to suitable registries to deploy in cluster.
Deploy the resources as below based on managed and unmanaged clusters.
Review deploy/ibmCloud/kustomization.yaml
file.
- Update images if required
- name: cos-driver-image
newName: icr.io/ibm/ibm-object-csi-driver
newTag: v0.1.11
- Update IBM COS endpoint and locationconstraint as per the region of your cluster
value: "https://s3.direct.au-syd.cloud-object-storage.appdomain.cloud"
value: "au-syd-standard"
- If you want to have 1-to-1 mapping between each PVC and secret(using same name for both), uncomment these lines in
deploy/ibmCloud/kustomization.yaml
file. This will automatically handle secret-name and namespace details, so you won’t need to add annotations for secret-name and namespace in the PVC configuration.
- target:
kind: StorageClass
patch: |-
- op: replace
path: /parameters/csi.storage.k8s.io~1node-publish-secret-name
value: "${pvc.name}"
- op: add
path: /parameters/csi.storage.k8s.io~1provisioner-secret-name
value: "${pvc.name}"
- op: add
path: /parameters/csi.storage.k8s.io~1provisioner-secret-namespace
value: "${pvc.namespace}"
Note: By default, in the IBM Object CSI Driver, the secret name is not tied to the PVC name. This allows you to use a single secret across multiple PVCs. For this, you’ll need to add two specific annotations in the PVC YAML. These annotations help the driver map the PVC to the correct secret.
annotations:
cos.csi.driver/secret: "custom-secret"
cos.csi.driver/secret-namespace: "default"
Deploy the driver
kubectl apply -k deploy/ibmCloud/
To clean up the deployment
kubectl delete -k deploy/ibmCloud/
After deployment following storage classes will be available in the cluster
NAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE ALLOWVOLUMEEXPANSION AGE
ibm-object-storage-smart-rclone cos.s3.csi.ibm.io Delete Immediate false
ibm-object-storage-smart-rclone-retain cos.s3.csi.ibm.io Retain Immediate false
ibm-object-storage-smart-s3fs cos.s3.csi.ibm.io Delete Immediate false
ibm-object-storage-smart-s3fs-retain cos.s3.csi.ibm.io Retain Immediate false
ibm-object-storage-standard-rclone cos.s3.csi.ibm.io Delete Immediate false
ibm-object-storage-standard-rclone-retain cos.s3.csi.ibm.io Retain Immediate false
ibm-object-storage-standard-s3fs cos.s3.csi.ibm.io Delete Immediate false
ibm-object-storage-standard-s3fs-retain cos.s3.csi.ibm.io Retain Immediate false
First review deploy/ibmCloud/kustomization.yaml
file and update IBM COS endpoint and locationconstraint as per the region of your cluster
value: "https://s3.direct.au-syd.cloud-object-storage.appdomain.cloud"
value: "au-syd-standard"
Then review deploy/ibmUnmanaged/kustomization.yaml
and update images if required
- name: cos-driver-image
newName: quay.io/containerstorage/ibm-object-csi-driver
newTag: v0.1.11
Deploy the driver
kubectl apply -k deploy/ibmUnmanaged/
To clean up the deployment
kubectl delete -k deploy/ibmUnmanaged/
Provide proper values for parameters in secret under examples/kubernetes/cos-s3-csi-<mounter_type>-secret.yaml
-
Create Secret, PVC and POD Pick the respective files based on the mounter you want to deploy. Below is for s3fs mounter.
kubectl create -f examples/kubernetes/cos-s3-csi-s3fs-secret.yaml
If you want to use your own bucket, bucketName should be specified in the secret. If left empty, a temp bucket will be generated.
kubectl create -f examples/kubernetes/cos-s3-csi-s3fs-pvc.yaml
kubectl create -f examples/kubernetes/cos-csi-app-s3fs.yaml
For rclone mounter, if any other rclone mount options need to be provided, they can be passed in Secret using StringData field. For example -
stringData: mountOptions: | upload_concurrency=30 low_level_retries=3
For non-root user support, in the Secret user can add
uid
which must matchRunAsUser
in Pod spec. Example -stringData: uid: "3000" # Provide uid to run as non root user. This must match runAsUser in SecurityContext of pod spec.
User can skip changes in Secret and directly use Pod Spec to enforce non root volume mount by providing
RunAsUser
value same asFsGroup
. -
Verify PVC is in
Bound
state -
Check for successful mount as below:
If mounter type is rclone
, verify using command
# mount | grep rclone
rclone-remote:rcloneambfail on /data type fuse.rclone (rw,nosuid,nodev,relatime,user_id=0,group_id=0)
If mounter type is s3fs
, verify using command
# mount | grep s3fs
s3fs on /data type fuse.s3fs (rw,nosuid,nodev,relatime,user_id=0,group_id=0,allow_other)
Collect logs using below commands to check failure messages
oc logs cos-s3-csi-controller-0 -c cos-csi-provisioner
oc logs cos-s3-csi-driver-xxx -c cos-csi-driver