[PVE-User] Unable to provision a virtual machine on local storage by cloning a template from a shared storage

[Ronny Aasen]

On 30.07.2019 20:05, Ntrs Ntrs wrote:
> Hi
>
> We have 2 nodes (Prox1 and Prox2) in a cluster. Each node has local-lvm LVM for KVM guests and shared NFS mount for templates available on both nodes.
>
> Prox1 storage:
>   - Local LVM
>   - shared NFS mount
>
> Prox2 storage:
>   - Local LVM
>   - shared NFS mount
>
> The templates are located on Prox1 on the NFS storage.  When cloning a template from Prox1 to Prox2 the only option for 'Target storage' is the shared NFS storage, which is a deal-breaker for us.
>
> Is it possible to clone a template from one node to another if the destination storage is not shared (i.e. not available on the source node)? Are we doing something wrong?
>
> In our case we want to clone a template from Prox 1 stored on the shared NFS mount to Prox 2 on the local LVM. This is required in order to use commercially available billing software that lists the templates from a single common location (the shared NFS mount) on the order form and provisions the KVM virtual server on the first available node on it’s local LVM storage.


just a check... since you have only 2 nodes I assume you have found a 
different way to deal with the split brain problem of no quorum, when a 
node dies/restarts ? your cluster will of course crash and die on a node 
reboot if you have not.


If the template is on non-shared storage on a node, it can only be 
cloned to storage on that node. If the template is on shared storage, it 
can be cloned to shared storage on any node.
so you can have the template as qcow2image on shared nfs you can clone 
it to a  vm on the shared nfs on any node. once the vm is up you can 
live migrate the local if you need that.

i do not know the software you talk about. but perhaps you can do some 
tricks like.. not have a prox template, only the disk image on nfs and 
instantiate a new vm with a copy of that disk image. or perhaps have the 
same template as a copy on each node using the same source disk.


good luck

Ronny