I am researching Virtualization/Private Cloud management tools before narrowing the list of tools to evaluate. The tool must be able to support our existing environments. This is the one criteria that I have not been able to verify from my reading of OpenNebula docs thus far.
The RDM datastore simply passes the device path to hypervisor. This <**source dev='/dev/disk/by-id/scsi-3xxasomestring here01'/**>* is what it is gathered from the PATH variable in the image template.
Per the docs “The RDM Datastore is an Image Datastore that enables raw access to block devices on Nodes. This Datastore enables fast VM deployments due to a non-existent transfer operation from the Image Datastore to the System Datastore.”.
Given the above, if ALL images for the OS’s and their data will be RDM’s, do we still need a SYSTEM datastore? If yes, how will it be used?
where
deployment.0 is the config file for libvirt (as would usually be found in /etc/libvirt/qemu/<vm_name>.xml
disk.0 is a sym link to the image/lun in the RDM datastore assigned to the VM (raw device/OS disk)
disk.1 is an ISO 9660 CD-ROM image containing a single script, context.sh, that sets variable values such as ETH0_MAC, TARGET=‘hda’ etc
ds.xml contains the config for the system datastore - onedatastore show <ds_name>
vm.xml contains the vm config - onevm show
The ISO is the contextualization service used to configure the guest OS. The variables set in the Context section on the VM Template are copied to the context.sh fiel to then be used by a daemon that runs in the guest OS provided it has the contextualization package installed.