On Tue, Apr 19, 2016 at 09:19:02PM +0800, Fam Zheng wrote: > On Tue, 04/19 14:07, Richard W.M. Jones wrote: > > We've done this successfully for years, for people monitoring their > > VMs using virt-df, pulling out files using guestfish and so on. We > > allow you to do it while the guest is live and running, with the > > proviso that a consistent view cannot always be guaranteed (although > > it works so reliably that it's never really a problem), or users can > > briefly pause VMs if they need a guaranteed consistent view. > > > > I'm afraid the onus is on you to explain why this existing practice is > > a bad idea. > > It is bad idea because it can produce erroneous data. Perhaps it's not > critical and is rare enough to be practically useful.
As explained above, we deal with the inconsistent case (by detecting and ignoring it, or retrying), and then there's the case where we pause the guest to get consistent data. > As a tradeoff, I guess, we can skip the shared lock in this series. Does that > work for you? I'd prefer some kind of no lock / ignore lock. There is a legitimate case where you want to have the shared lock behaviour, but also a legitimate one for turning it off. I'm not opposed to the idea -- there are very real cases where your patch saves people from themselves. Off topic: How does this patch deal with genuinely shared (writable) disks, as used in cluster filesystems? Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-df lists disk usage of guests without needing to install any software inside the virtual machine. Supports Linux and Windows. http://people.redhat.com/~rjones/virt-df/
