On Tue, Sep 30, 2014 at 10:09:36AM -0400, Matthew Miller wrote:
This may be related to the automatic partition resizing, but I'm not sure.
PVM-based booting, as in EC2 (uh, except for HVM in EC2, obviously) is
We should probably add "reboots properly in HMV environment; and reboots
properly in PVM environment" to our validation tests and release criteria.
I noticed something similar to this a while back  but I assumed it was either
because I was using pre-alpha grade images or it was because I was using virt-install
to do the image build process.
Basically what I noticed is that doing the image "build" using the same tree/etc
koji yielded me with a VM that would boot once (so anaconda install and first boot work
fine). However, subsequent boots would fail. I would then have to dd the syslinux
mbr.bin to vda and everything would work fine.
After looking at  I assumed this was a conscious decision we had made to not
install the mbr by default (although reading through the comments now shows that you
are only affected by this if the disk was not blank prior to build) and for some
reason I assumed that it just worked on cloud platforms (although at the time I
didn't have an image to test).
Long story to say that I agree it looks like something on boot is affecting the mbr.
 - https://lists.fedoraproject.org/pipermail/cloud/2014-August/004166.html
 - https://bugzilla.redhat.com/show_bug.cgi?id=1015931