Testing draft for vsphere/photon volume plugin/cloud provider

[luomiao]

1. K8S unit test make test WHAT="pkg/volume/photon_pd pkg/cloudprovider/providers/photon" make test WHAT="pkg/volume/vsphere_volume pkg/cloudprovider/providers/vsphere"

2. Storage specific tests a. Basic volume plugin tests

• [ ] Back-to-back pod creation/deletion with the same volume source on the same worker node - [vsphere completed]
• [ ] Same volume source attach/detach to different worker nodes - [vsphere completed]
• [ ] Back-to-back pod creation/deletion with different volume sources - [vsphere completed]
• [ ] Data correctness check for volume source reuse (this one can be part of other tests) - [vsphere completed]
• [ ] Filesystem type test - [vsphere completed]
• [ ] Test multiple volumes of same datastore within the same POD - [vsphere completed]
• [ ] Test multiple volumes on different datastores within the same POD - [vsphere completed]
• [ ] Test combination of PV, PVC and DP in same POD - [vsphere completed]
• [ ] Service comprising multiple PODs where each POD uses his own volume - [vsphere completed]

• [ ] E-2-E App including WebServer, Database using DP

  b. PV/PVC tests

• [ ] Static Provisioning: PV Creation/PVC Binding/Pod Creation/Pod Deletion/PVC Deletion/PV Deletion - [vsphere completed]
• [ ] Dynamic Provisioning: Storage Class Creation/PVC Creation/Pod Creation/Pod Deletion/PVC Deletion/Storage Class Deletion - [vsphere completed]
• [ ] Reclaiming Policy Test (Only "deletion" is supported) - [vsphere completed]
• [ ] AccessMode Test (Only "RWO" is supported?)

• [ ] Label Selector Test - [vsphere completed]

  c. Volume source usage in Replication Controller TBD

  d. Volume source usage in StatefulSet (PetSet) TBD

  e. Failover tests

  • [ ] Node failure: Does POD restart successfully with volume accessible on new node?
  • [ ] POD failure: Does POD restart successfully with volume accessible on new node?
  • [ ] vCenter/PC failure: Existing services should continue to work and provisioning should fail with graceful error message.
  • [ ] Cluster failure: Volume should be persistent after the whole K8S cluster is deleted [Balu] - True only if disk is detached from nodes. If not, all the volumes would be deleted.

3. Kubernetes-anywhere tests

• [ ] Node restart test: all required K8S services should be up
• [ ] Kubernetes-Anywhere should be able to find template in different resource pool and folder.
• [ ] Kubernetes-Anywhere should be able to launch VMs in specified VC cluster, resourcepool.
• [ ] Kubernetes-Anywhere should be able to launch cluster from master branch.

[luomiao]

where we should hold these testing scripts? option 1: vmware/kubernetes with a new branch option 2: luomiao/testing-scripts option 3: vmware/vmware-private-kubernetes-collaboration other recommendations are welcome.

[kerneltime]

Filed https://github.com/vmware/kubernetes/issues/50 to track investigation on how to implement and run tests.

[kerneltime]

Stateful sets

• [ ] Stateful Set consisting of multiple copies

• [ ] Scaling up and down stageful sets

• [ ] Test if a names instance of a pod on recreation gets the same volume

  Resource Quota tests

• [ ] Total size limitation across all claims

• [ ] Total size limits for a storage class

• [ ] Total count limits on claims

• [ ] Total count limit for a storage class

Negative tests

• [ ] Volumes claimed by a namespace are not available on a different namespace