Skip to content

Commit b7bfaa0

Browse files
committed
Update Readme with more details and considerations
1 parent 2efecfc commit b7bfaa0

File tree

1 file changed

+11
-0
lines changed

1 file changed

+11
-0
lines changed

README.md

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -80,6 +80,11 @@ The storage class must also have a parameter named
8080
`csi.cloudstack.apache.org/disk-offering-id` whose value is the CloudStack disk
8181
offering ID.
8282

83+
**Reclaim Policy**: Storage classes can have a `reclaimPolicy` of either `Delete` or `Retain`. If no `reclaimPolicy` is specified, it defaults to `Delete`.
84+
85+
- `Delete`: When a PVC is deleted or a CKS cluster (Managed Kubernetes Cluster in CloudStack) is deleted, the associated persistent volumes and their underlying CloudStack disk volumes will be automatically removed.
86+
- `Retain`: Persistent volumes and their underlying CloudStack disk volumes will be preserved even after PVC deletion or cluster deletion, allowing for manual recovery or data preservation.
87+
8388
#### Using cloudstack-csi-sc-syncer
8489

8590
The tool `cloudstack-csi-sc-syncer` may also be used to synchronize CloudStack
@@ -223,6 +228,12 @@ kubectl logs -f <cloudstack-csi-controller pod_name> -n kube-system # defaults t
223228
kubectl logs -f <cloudstack-csi-controller pod_name> -n kube-system -c external-provisioner
224229
```
225230

231+
## Additional General Notes:
232+
233+
**Node Scheduling Best Practices**: When deploying applications that require specific node placement, use `nodeSelector` or `nodeAffinity` instead of `nodeName`. The `nodeName` field bypasses the Kubernetes scheduler, which can cause issues with storage provisioning. When a StorageClass has `volumeBindingMode: WaitForFirstConsumer`, the CSI controller relies on scheduler decisions to properly bind PVCs. Using `nodeName` prevents this scheduling integration, potentially causing PVC binding failures.
234+
235+
**Network CIDR Considerations**: When deploying CKS (CloudStack Kubernetes Service) clusters on pre-existing networks, avoid using the `10.0.0.0/16` CIDR range as it conflicts with Calico's default pod network configuration. This overlap can prevent proper CSI driver initialization and may cause networking issues within the cluster.
236+
226237
## See also
227238

228239
- [CloudStack Kubernetes Provider](https://github.com/apache/cloudstack-kubernetes-provider) - Kubernetes Cloud Controller Manager for Apache CloudStack

0 commit comments

Comments
 (0)