I have deployed OpenNebula 4.14 and wonder if and how it is possible to deploy the disk images of a VM to more than one system datastore.
This is my scenario: We have setup 2 Gluster FS storage clusters. One big one that is built on SAS disks and a smaller one built on fast SSDs. The clusters are visible in OpenNebula as two system datastores. For VMs running applications with higher performance requirements I would like to locate one smaller disk image on the SSD-based system datastore for data that is frequently read and written. The bulk of the data I would like to store on a disk image located on the 2nd slower system datastore.
hi, I tried this before (using 4.12) - OS image on local disk and the data disk from remote NFS.
so VDA was on datastore 1 and VDB was on datastore 2, as long as both datastores are useable and monitored, that should work.
When editing the template, under tab called “storage”, add the first disk from the datastore. Then click “add another disk” and select the second disk from the other datastore. save & deploy VM
Thanks for your hints. Unfortunately in my case this does not seem to work. During deployment the 2 disk images are drawn from the same image datastore and are placed in the same system DS. I don’t find a possibility to define on a per disk basis which system datastore shall be selected. Also the DS placement policies only seem to work on a per VM basis.
The SYSTEM datastore is the “home” of the VMs. A VM can’t have two home directories. When using shared/clustered/ storage for the images you should define them as IMAGES data stores with proper DS_MAD and TM_MAD. Your target is to have only symlinks to the disk files on the desired storage clusters in the VM’s SYSTEM datastore.
thanks for the clarification. I guess I missunderstood some of the Open Nebula concepts. I will try out this as soon as possible.