we used Amazon AWS for a long time and want to build a private cloud for our needs. On the way I found Open Nebula and find it easy to used and to install compared to Open Stack.
Now I like to archive a “simple” thing in my eyes and do not know how to do that in Open Nebula 5.x.
On one node I have two raids one on magnetic discs and one on ssd discs. I like to make attachable images for an instance. One image for the OS, one image on a magnetic discs and one image on ssd discs for timecritical data, like webapps.
Like this (/home/one/datastores is the default path defined in the config files):
|------> OS image (sda) on /home/one/datastores
|-------> Data image (sdb) on /home/one/datastores
|-------> Data image (sdc) on /home/one/datastores-ssd
In Open Nebual 4.x I could define a BASE_PATH to a datastore. There I could define the path to the magnetic disc array and to the ssd disc array.
How can I accomplish that with Open Nebula 5.x? Best would be to do that on Sunstone directly.
Thanks in advance,
BASE_PATH was generating to much troubles, in terms of SOURCE paths for VMs
so we decided to fix the path location (more details here:
http://dev.opennebula.org/issues/4316) . Now you can create the datastores
and symlink them like:
Datastore 100 (ssd)
/home/one/datastore/100 —> /mnt/dastore-ssd/100
Datastore 101 (hd)
/home/one/datastore/101 —> /mnt/dastore-hd/101
So this is more a setup issue now
thanks for your reply. I understand the decision. Makes sense to me.
I did exactly that, but I have to use a shared datastore for that. Otherwise the images will be copied from the /home/one/datastore directory to the VMS directory. Meaning that the images are in the SYSTEM datastore.
For the time being until I find a better solution I solved the issue making each node a cluster too. Setup datastores in each cluster for each purpose (System, File and Images).
The negative aspect of this is that I need to download all Applicants/Templates to every cluster. I liked the idea to have the Applicants/Templates on the node where Sunstone is running and let them be copied (SSH datastore). But with that all VMS stuck in the state pending. It seems not work over cluster boundaries.
Other Point is that the list of datastores and images getting crowded in Sunstone…
I like to have the applicants and images from the default cluster with SSH datastores from the frontend node be copied to machines that a sitting in a different (node only cluster), while I am using the node only cluster to setup the datastores as shared datastores, to be able generating an image on an SSD drive and link it to the VM while other images for the same VM are on the magnetic drive.
I’m not getting this one. The idea is to have each datastore directory symlinked to the correspodint storage backend. /home/one/datastore is just the path for all the datastores either system or images. i.e. the BASE_PATH. But does not need to be shared; you just need to setup the links and structure in the hypervisors.