On Wed, Oct 16, 2013 at 02:24:59PM -0600, Chris Murphy wrote: > > On Oct 16, 2013, at 1:50 PM, Richard W.M. Jones <rjones@xxxxxxxxxx> wrote: > > > On Wed, Oct 16, 2013 at 01:09:25PM -0600, Chris Murphy wrote: > >> Caching mode was none in all prior cases. > > > > Note that cache=none is almost never useful. It bypasses the host > > cache so you don't get the benefit of having lots of free host memory, > > plus causes endless pain because of filesystems that don't support > > O_DIRECT (eg. tmpfs) or have buggy O_DIRECT support (eg. glibc on top > > of older kernels). It took me quite a long time to realize this -- it > > was the default in libguestfs until quite recently :-( > > Curious, it seems [none] should perform almost as well as > writeback, No, because writeback uses the host memory as a very big cache. Whereas none can only use the guest's memory, which is generally much smaller than the host's memory. Bigger cache = better performance, all else being equal. Note that writeback is only safe if your guest sends flush commands. In your case this is true because you are installing new Linux guests. But it would not be true if you were testing, (eg) RHEL 5.0 with LVM. > and much better than writethrough. Correct, because writethrough forces every write to disk. This is only needed for guests that don't send flush commands. > But safer than either writeback or unsafe. Writeback is safe, provided your guest sends flush commands. Unsafe deliberately ignores guest flushes in the name of speed. > I think the qcow2 > metadata preallocation is what's making a bulk of the difference in > what I'm seeing. Yes, quite likely this makes a substantial difference. virt-v2v got a big boost when we started to use preallocation=metadata. > So is a compromise between performance and some safety to use > writeback instead of unsafe? Yes. Libguestfs (now) lets you switch between writeback and unsafe, and doesn't let you choose any of the other cache modes. The reasoning for this is described here: https://github.com/libguestfs/libguestfs/commit/749e947bb0103f19feda0f29b6cbbf3cbfa350da Note this reasoning only applies because with libguestfs we can be sure that we have a modern kernel which sends flush requests. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones libguestfs lets you edit virtual machines. Supports shell scripting, bindings from many languages. http://libguestfs.org -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
