My host servers have lots of local storage: 6 x 1TB disks. I want to use that local storage, so I have local storage configured "on" for non-system VMs.
But when I add a host to CloudStack management, it configures one local storage pool on that host, located under /var/lib/virt/images in the "/var" volume -- and on my systems /var is only 8GB in size! So I need to configure 6 new local storage pools per host, one for each disk. Neither the GUI nor the API seem to support adding local storage pools: - In the GUI, under Infrastructure > Primary Storage > +Add Primary Storage > Protocol, there are only shared storage protocols allowed (NFS, SharedMountPoint, RBD, and CLVM). - In the API, CloudMonkey and the API docs, under "create storagepool" the allowed arguments are only zoneid/podid/clusterid, name, "details", and url, which do not seem to allow for local storage either. THIS CAN BE DONE: (hack alert) I went to a lower level and looked at libvirt and virsh on the host server. I was able to use virsh to analyze the existing storage pools and create the needed new "dir"-type local storage pools on the host (details available if desired), but the pool did not get registered in CloudStack although I restarted the agent and then the manager. Still, I was able to substitute one of the new 1TB volumes for the original cloudstack-defined 8GB volume by the following gross hack: a) record the UUID of the cloudstack-defined volume b) put the host in maintenance mode and stop the cloudstack-agent c) virsh pool-destroy both its volume and my volume locally on the host d) virsh pool-create a new volume in the 1TB location with the cloudstack-defined UUID and name e) restart the cloudstack-agent and take the host out of maintenance mode It popped an exception in the logs and an error in the web app, but successfully finished taking the host out of maintenance mode. This host is now happily hosting CloudStack VM instances with big local-store Root volumes. So, there must be a way to use MySQL to edit the database and insert these additional local storage pools into the system (and take out the /var pools that are too small to do any good). Would someone please tell me what to put where, so I don't have to reverse engineer the many storage-related database tables? I'm using CloudStack 4.1.0 over CentOS 6.3 with KVM. Thanks, greatly appreciate any help, --Matt
