Red Hat Bugzilla – Bug 476296
Paravirtualized SLES10SP2 install crashes during DHCP
Last modified: 2010-09-23 10:30:44 EDT
Created attachment 326780 [details]
Description of problem:
When I install a SLES 10.2 Guest OS it doesn't make it all the
way through the installation process. Right afer I select DHCP as the IP acquisition method the VM shutsdown. I've attached some screen shots of that as well as the virt-manager.log.
Version-Release number of selected component (if applicable):
RedHat 5.3 snapshot 5
any time SLES 10.2 guest OS is installed.
Steps to Reproduce:
1. Install a paravirtualized SLES 10.2 guest OS
2. During the installation process select NFS as the install method
3. and slect DHCP as the IP acquisition method
(I don't know if this is necessary, but that is how I ran into the issue.)
Created attachment 326781 [details]
Installation CD not Found Image
Created attachment 326782 [details]
Use DHCP image
Right after this screen shot is when the VM reboots.
Since I can get this to work under QEMU on F9, I'm going to assume this is a xen bug, and outside the realm of virtinst. Re-assigning to xen.
(In reply to comment #4)
> Since I can get this to work under QEMU on F9, I'm going to assume this is a
> xen bug, and outside the realm of virtinst. Re-assigning to xen.
Cole, were you able to run it under KVM (not QEMU without KVM) ? There were some issues with running something under KVM and Xen but not on QEMU without KVM in the past. I think it was about Minix.
Don't really remember to be honest. I'd see if KVM in RHEL5 has the same issue.
Ok, good Cole. I don't know. However, what location could I download PV images of SLES10 SP2 install?
could you please tell me the location to install SLES10 SP2 from?
We have a local NFS repository that we use to install SLES from. I'm not sure where you would be able to get a copy of SLES for yourself. Do you typically test with SLES VMs?
honestly no. I've never worked/tested with SLES VM yet so the answer's no. I need to try but from what I know it's not available from their site for evaluation anymore so this may be hard to find.
Just a clarification: There's SLES 11 SP1 for evaluation download but not SLES 10 SP2 anymore. Do you think it's ok to test with SLES 11 SP1 or isn't it an issue there? What do you think? Do you have some experience with SLES 11 SP1 testing and whether this issue is present or not?
Honestly I'm not sure. If you have access to the SLES 11 SP1 it wouldn't be a very long test to try to create a VM with SLES 11 SP1. If the OS installs successfully then the problem doesn't exist with SLES 11 SP1. If it doesn't, then you can use that test bed for debugging.
Hector, this is only in the form of ISOs and not the images so it won't work to install the PV guest AFAIK.
You can always mount the ISOs and copy their contents to a local directory. (It would work for Fedora at least, likely SLES is not too different).
(In reply to comment #17)
> You can always mount the ISOs and copy their contents to a local directory.
> (It would work for Fedora at least, likely SLES is not too different).
Oh, so are you saying it could work when I mount ISO image to dom0 and provide the patch to /media/cdrom/os or something like that?
I can try. Also, I think I found something relevant in upstream c/s 21808  since this is having some note about SLES 10 SP1 so I guess this may be some other error but it may be relevant.
Thanks a lot Paolo, I didn't know that this would work for PV guests,
 http://xenbits.xensource.com/staging/xen-unstable.hg?rev/c4a83e3cc6b4 (please read the commit message there)
Well, I did try it but unfortunately I there are many issues with SLES 11 SP1 x86_64 installation. I was able to install it as HVM guest but not as PV guest. The problem is that PV guest installation requires a CD-ROM which is not available for PVs at all so therefore I used the ISO to install it as PV guest as described and advised for SLES 10 SP2 PV guest at  on the official Novell site, then I tried reconfiguring the guest to be a PV guest too but there are many issues. I installed the virtualization host to install the xen kernel and then I edited boot/grub/menu.lst to include the xen kernel there.
The problems are mainly because of missing PV drivers and therefore disk and NIC devices inaccessible.
 http://www.novell.com/communities/node/3046/how-boot-paravirtual-xen-domu-cd-rom-or-iso-sles-10-sp1-or-opensuse-103 - when reading it I guess the article was relevant to this one
The patch is interesting, but installation fails for the reporter much earlier than when pygrub is run.
> The problem is that PV guest installation requires a CD-ROM which is not
> available for PVs at all
What do you mean?
(In reply to comment #20)
> The patch is interesting, but installation fails for the reporter much earlier
> than when pygrub is run.
Well, I was unable to run into those issues at all. In fact the only thing I was able to do was to install HVM guest because when I was installing it as PV guest I run into issues that installation *always* told me that I need to have a SLES CD inserted in the CD-ROM drive.
> > The problem is that PV guest installation requires a CD-ROM which is not
> > available for PVs at all
> What do you mean?
Like I wrote above, when running the SLES 11 x86_64 PV guest installation I have no choice of installation media and it keeps asking me to insert the SLES CD-ROM to the CD-ROM drive but for PV guests there is no CD-ROM drive support so I can't install it at all.
Note that this is about SLES 11 x86_64 but the bug was reported against SLES 10 SP2 which is no longer available for evaluation so I'm afraid I can't run into the same issues like described in comment #0 because of SLES 11 may have changed in some respects.
PV guests do support a CD-ROM, though not a "passthrough" of a *physical* CD-ROM. In other words, you can only add a iso image as a CD-ROM, but that *is* possible. You can "change" it by detaching the device and reattaching it.
xm block-attach <domain> file:/path/to/cd1.iso hdc:cdrom r
xm block-detach <domain> hdc
xm block-attach <domain> file:/path/to/cd2.iso hdc:cdrom r
This has been stalled for a while (2 months), and the last bugzilla comments are not really relevant for the bug.
So, to sum up:
1) Is there any way we can get images? (e.g. we can try asking SEG?)
2) Or maybe the reporter could try again and gather an sosreport of the host or at least /var/log/messages, /var/log/xend.log and "xm dmesg", which is standard procedure for all bug reports?
If we lack both of these items, the only sensible choice is CLOSED/INSUFFICIENT_DATA.
As I work in the team that originally filed this, my recommendation would be to close this BZ. We are not strategically supporting RHEL 5 Xen any longer.
Virtualization and Linux Pgr. Mgr.