Open Stack

Hi All!I have recently been troubled by qemu's disk cache, I learned from here the difference between them.And I think that writethrough should be able to guarantee that the virtual machine will not have data loss in the case of a host power outage.But my test results show that writethrough does not guarantee data security, only in the case of dd with oflag = sync(no matter writeback/none/writethrough), to ensure that the data is not lost. So I think that the different modes of qemu disk cache just guarantee that the data is missing less, writethrough does not mean that we can be assured that the data will not be lost.Is my understanding right? Or are I totally wrong?
about environment
qemu version: 2.3.1openstack version: junohost os: CentOS 7guest os: CentOS 7 and close lvm write cache (/etc/lvm/lvm.conf write_cache = 0)Cinder backend storage: IPSANTest method:    guest os: dd if=/dev/zero of=/var/Test_disk bs=1M count=1    host os: kill -9 ${guest_pid}
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20170619/2b916fbd/attachment.html>