On 06/13/2012 01:56 PM, Charles Chou wrote:
Hi,
After reading the source some more I think iscsi is a top
level option so you may want to try:
-iscsi initiator-name=some_name -drive
file=iscsi://[xxxx:xxxx:xxxx::3]:3260/iqn.2010-08.xx.xxx.xxxxxxx.pool10-images/10,....
Also I am curious as to whether it performs better than using
virtio on a block device that is created by the open-iscsi
initiator.
this is going off the original thread topic for a moment. i wouldn't
expect the performance to be any better. my interest in using the
iscsi qemu block backend is that it can be done without modifying
the host. you don't have to have escalated permissions to connect
the host kernel to the iscsi target or figure out which block device
matches up to the iscsi target and lun or have permissions to
read/write from the kernel block device. so it adds some separation
between the guest and the host kernel. (though i brought this up
with the ovirt guys last fall and got lots of 'wtf' looks)
mike
|