I am reading about implementing the unified NetApp Cinder driver for OpenStack, but have a question about the documentation. Specifically the part in bold.
I am using Clustered Data Ontap 8.3.
netapp_vserver = None
(StrOpt) This option specifies the virtual storage server (Vserver) name on the storage cluster on which provisioning of block storage volumes should occur. If using the NFS storage protocol, this parameter is mandatory for storage service catalog support (utilized by Cinder volume type extra_specs support). If this option is specified, the exports belonging to the Vserver will only be used for provisioning in the future. Block storage volumes on exports not belonging to the Vserver specified by this option will continue to function normally.
Does this mean that the SVM that you specify here is not usable anymore from outside the OpenStack environment? I am a bit hesitant because I also need to specify the cluster management login details.
"Does this mean that the SVM that you specify here is not usable anymore from outside the OpenStack environment? I am a bit hesitant because I also need to specify the cluster management login details."
No, the SVM is still "usable" but the creation of files, LUNs, etc will be managed by OpenStack. The ideal mapping is 1 Cinder Backend to 1 SVM/Vserver.
The scenario is that there is a SVM for development, where there is one volume that is meant for OpenStack (PoC). The SVM isn't solely for this OpenStack environment, but also needs to be managed using the NetApp OnCommand system to create and manage other volumes that have nothing to do with OpenStack.
So as the SVM is managed by OpenStack, I should rather make a dedicated SVM for OpenStack?
I am not sure if I understand entirely how this works: I thought there was a volume (or multiple) that I gave to OpenStack (cinder) and cinder creates multiple cinder-volumes that use the available space of the volume NetApp delivers. So why is there even a need for SVM management?