How to create an OpenNebula datastore

OpenNebula datastores define where VM images, running VM disks, and file artifacts live before the scheduler can place workloads on hosts. Creating another datastore is useful when a cluster needs a separate storage pool, a shared image repository, or a capacity boundary away from the default datastore.

The CLI creates a datastore from a small template file. A shared filesystem image datastore backed by an NFS export is a common starting point because DS_MAD="fs" stores images as files, TM_MAD="shared" tells OpenNebula that hosts see the same datastore path, and TYPE="IMAGE_DS" makes the datastore available for image registration.

The storage export must already be reachable by the OpenNebula front-end and by the hosts in the target cluster. Match the transfer driver with compatible system datastores before scheduling VMs, and keep the smoke image small so the final check proves write access without consuming useful capacity.

1. List the existing datastores before choosing the new name and transfer mode.

   $ onedatastore list
     ID NAME                SIZE AVA CLUSTERS IMAGES TYPE DS      TM      STAT
      2 files                50G 86% 0             0 fil  fs      local   on
      1 default              50G 86% 0             2 img  fs      local   on
      0 system                 - -   0             0 sys  -       local   on

2. Create the datastore template file.

   nfs-images.ds

   NAME            = "nfs-images"
   DS_MAD          = "fs"
   TM_MAD          = "shared"
   TYPE            = "IMAGE_DS"
   NFS_AUTO_ENABLE = "yes"
   NFS_AUTO_HOST   = "storage.example.net"
   NFS_AUTO_PATH   = "/exports/opennebula/images"
   NFS_AUTO_OPTS   = "soft,intr,rsize=32768,wsize=32768"

   Replace the NFS host and export path with the storage server that the front-end and hypervisor hosts can reach. For a system datastore, use TYPE="SYSTEM_DS" and omit DS_MAD; for a files and kernels datastore, use TYPE="FILE_DS" with the driver pair required by that storage design.

3. Create the datastore from the template.

   $ onedatastore create nfs-images.ds
   ID: 100

4. Inspect the new datastore state and template attributes.

   $ onedatastore show nfs-images
   DATASTORE 100 INFORMATION
   ID             : 100
   NAME           : nfs-images
   USER           : oneadmin
   GROUP          : oneadmin
   TYPE           : IMAGE
   DS_MAD         : fs
   TM_MAD         : shared
   BASE PATH      : /var/lib/one//datastores/100
   STATE          : READY
   
   DATASTORE TEMPLATE
   DS_MAD="fs"
   NFS_AUTO_ENABLE="yes"
   NFS_AUTO_HOST="storage.example.net"
   NFS_AUTO_PATH="/exports/opennebula/images"
   TM_MAD="shared"
   TYPE="IMAGE_DS"

5. Add the datastore to the target cluster.

   $ onecluster adddatastore production nfs-images

   A datastore can belong to more than one cluster, but every host in those clusters must be configured for the same storage path and transfer behavior.
   Related: How to create an OpenNebula cluster

6. Verify that the cluster lists the datastore ID.

   $ onecluster show production
   CLUSTER 100 INFORMATION
   ID             : 100
   NAME           : production
   
   HOSTS
   0
   
   VNETS
   1
   
   DATASTORES
   100

7. Create a small test image in the new datastore.

   $ oneimage create --name datastore-smoke --type DATABLOCK --size 64 --datastore nfs-images
   ID: 42

   The test image allocates a small datablock only to prove that OpenNebula can create an image in the datastore.

8. Check that the test image is ready in the datastore.

   $ oneimage show 42
   IMAGE 42 INFORMATION
   ID             : 42
   NAME           : datastore-smoke
   USER           : oneadmin
   GROUP          : oneadmin
   DATASTORE      : nfs-images
   TYPE           : DATABLOCK
   STATE          : READY

9. Remove the test image after the write check.

   $ oneimage delete 42

   Delete only the temporary smoke-test image. Do not remove production images that were registered in the datastore.