Red Hat Bugzilla – Bug 612525
RHEL 6B2 install bombs out during graphical install on KVM
Last modified: 2013-01-09 17:50:31 EST
Created attachment 430341 [details]
a full tarball of the contents of the /tmp directory from the crashed server
Description of problem:
When I try to install RHEL6 beta 2 under RHEL 5.5 in a KVM then
Version-Release number of selected component (if applicable):
(SHA266 = 272c6fbf1a423e5bab6fddb101b6a4c96b57bf196fa2b26991f55202e0fa715e)
Each time tried.
Steps to Reproduce:
1. virt-install --name=rhel6b2-test2 --ram=1024 --vcpus=1 --os-type=linux --os-variant=rhel6 --accelerate --cdrom=install-cds/RHEL6.0-20100622.1-Workstation-x86_64-DVD1.iso --disk path=/var/lib/libvirt/images/rhel6b2-test2.img,size=10,sparse=false --vnc --force --noautoconsole
2. take defaults all the way through install except tell to take entire disk
3. get to select installation packages ; leave as desktop
4. press next
install exited abnormally [1/1]
system should install and work
* Text mode install seems to work okay (edit install command line add "nomodeset text" at end).
* The first time I ran this through the X server log showed the X server bombing out. This time round there is no such thing, but something on console 6 seems to have hung though I can bet back and forward to it.
* I still have the VMs live so can provide access to the server for MM to log in.
I don't see anything in your logs that indicates the nature of the problem. If you're receiving a traceback, can you also make sure /tmp/anaconda-tb-* is attached to this file? If it's a stack trace, can you attach a screenshot of tty1?
That's a complete dump of /tmp/ there was no more. Also looking at the first console there's nothing between the start of graphical install and the message install exited abnormally [1/1]
On the other hand the VM is still active so if you can suggest how I can trigger the traceback you want I can do that.
I attempted this install before and had a problem at more or less the same stage though the X server crashed fully. There is also no /tmp/anaconda-tb-* file there either. I'm attaching the /tmp files from that as well just in case they provide inspiration.
Incidentally I should point out that on this machine there are several other VMs, but RHEL 5 and Ubuntu running and they seem to have no problem.
Created attachment 430397 [details]
/tmp files from my first install attempt
I did a previous install. I'm not sure about the exact procedure used; approximately select workstation and full disk install; this is the set of /tmp files produced that time. In this case the X server actually crashed out fully. I suspect that it's just a different symptom of the same root cause, but can't be sure.
Again, this is all the files from /tmp but the VM is still running if you want e.g. a signal sent to some process or something.
This may be related to bug 610255
What happens if you increase the memory for your VM when doing a graphical install?
With 4096M of memory the install completes and appears successful.
*** Bug 613104 has been marked as a duplicate of this bug. ***
Libvirt doesn't have any bearing on anaconda using too much memory inside the guest. We merely create a KVM virtual machine with the requested amount of RAM. How the guest OS uses that RAM is its own decision.
It might be another incarnation of Bug 607650 - Install exited abnormally while installing packages. Adding Avi to the CC list. Reported can try the brew build on that bug
I agree it worth testing with the fix for bug 607650. I don't think there's a backport to rhel5 host yet but there is a rpm for rhel6 host availabile in bug 607650.
Putting in needinfo until we can get results back for comment #11 and #12.
Can we dupe this one against bug #607650?
*** This bug has been marked as a duplicate of bug 607650 ***