Hello Everyone, We're evaluating OpenNebula 3.2 in our environment and have a few questions:
* What is the most optimal (and recommended) approach to provide redundancy / HA for front-end processes (oned and sched). I've been looking into drbd-type approaches, however I'm sure others already have this in their environments, and I'm hoping someone can provide guidance on the most reliable / best approach to achieve this. * Is it possible to (easily) build an image using ONE? At the moment we use another VM manager to build a KVM image, convert the qcow2 file to raw and import into ONE. It would be easier to build the image using ONE. Although it seems possible, it seems like a disconnected process in making multiple templates, saving disk images, and managing (temporary) ISO media. * Is it possible to change a deployed VM (template)? To clarify specifically: assume a deployed VM, and the requirement to add another disk / or contextualization variable to an already deployed VM. I've noticed that by simply updating the template from which the VM was deployed from, does not change the already deployed VM. I would like to avoid deleting and re-provisioning VMs for minor template changes. * Is it possible to separate images/templates from VM data files? Currently everything located in the /var/lib/oneadmin home directory. Ideally we'd like to split the data associated with templates/images/ISOs, etc and the actual VMs we deploy. * Is it possible to have MULTIPLE VM_DIRs defined? We are considering the use of CLVM (see question below) or OCFS2 to provide shared volume / file system access. The question relates to data classification and scaling data store capacity. Expansion of an OCFS2 FS is not possible, and migration to a larger volume involves downtime. Deploying additional VM data stores will provide the scalability we're after. CLVM would also allow us to add additional physical devices to a logical volume. * I noticed that ONE comes with TM_LVM. Does this imply that CLVM is supported? Relating to the above question, can we have multiple volume groups defined for data classification purposes? Are there any other considerations / dependancies that are necessary to enable CLVM support under ONE? * How does ONE scale to support multiple zones? My understanding (based on documentation) is that ONE follows a manager of managers model when it comes to zones. A ONE front-end (master) instance can administer downlevel (slave) ONE front-end instances that themselves manage hypervisors. Assuming HA is possible, this leads me to believe that two (2) ONE Front-End servers are necessary for every zone, in addition to two (2) ONE front-end servers for the manager of managers function (master). Given this approach, sclability can dramatically increase cloud overhead (physical to virtual server ratio), especially for multiple datacenters (data center = 1 zone) hosting 3-6 compute nodes each (this equates to 5-8 physical hosts considering HA front-end servers). UNLESS.... There are ways / methods that are proven and would be supported under a C12G maintenance contract that we can virtualize these ONE front-end servers... Thank you. _______________________________________________ Users mailing list [email protected] http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
