[openstack-dev] dd performance for wipe in cinder

Jeremy Stanley fungi at yuggoth.org
Fri Oct 11 16:58:54 UTC 2013


On 2013-10-11 08:44:00 -0600 (-0600), Chris Friesen wrote:
> Is there a particular reason why you're overwriting the entire
> volume with zeros?
> 
> A simple way to delete the contents of a filesystem would be "rm -rf
> /path/to/directory"

You'd like to think that, but many (most?) filesystems only unlink
inodes on deletion and leave the data behind in their old blocks to
be overwritten by the processes to which they are subsequently
reallocated. A user with root-level access to a thick-provisioned
block device presented into their virtual machine can still read the
old data from it if it hasn't already been overwritten.
-- 
Jeremy Stanley



More information about the OpenStack-dev mailing list