XML Feeds
rhel6 xen domU on a CentOS 5 dom0
Spent a bit of time last evening and tried to get a rhel6 beta domU going on a CentOS 5.4 dom0 - everything seems to go fine except that the VM will not boot. The problem is that the rhel6 anaconda creates a /boot with ext4, which pygrub on CentOS and RHEL 5 will not be able to boot.
To Fix this, make sure you select the option to review partition and disk layout, and manually change /boot from ext4 to ext2. Everything should just work fine.
- KB
28/Apr/2010 05:36:28 am, 10 comments
Comment from: Shane Falco [Visitor]
LOL, where were you last night? I had to figure this out for myself. :DHopefully gitco will get an updated pygrub that can handle ext4, but for now I'll just continue to use ext3 in my VMs.
28/Apr/2010 @ 07:43
the Gitco packages have never worked for me, and to be honest, the stock Xen Dom0 support already included in CentOS-5 are plenty for what I need :)
28/Apr/2010 @ 07:48
Comment from: John Hodrien [Visitor]
It's not like having an ext2/3 /boot is a hardship. There's little if any benefit to having an ext4 /boot, and this doesn't in any way impact the filesystems you can use elsewhere.
29/Apr/2010 @ 10:18
John,Thats true. However, I just feel that the installer should be smart enough to leave behind a working system. Specially since the Xen strategy for Red Hat is focused around using RHEL5 as the host machine with RHEL4 and RHEL6 only targetting a Domu state.
- KB
29/Apr/2010 @ 10:27
Comment from: Shane Falco [Visitor]
John, the issue is that many people (including myself) like to build our Xen domains using a separate logical volume for xvda1 mounted as / and another lv for xvda2 as swap. This allows us to *very* quickly build domains as simply as lvcreate, mkfs, mount, untar, unmount. In this methodology there's no good reason to split off a separate /boot.
29/Apr/2010 @ 20:03
Comment from: hugo [Visitor]
Seems like a fixed problem in 5.5?* pygrub did not understand fourth extended (ext4) /boot partitions, and so was unable to paravirtualize guest domains. e4fsprogs-devel and ev4sprogs-libs packages are provided with this update for pygrub and other applications that require the new ext4 capable e2fsprogs libraries. (BZ#528055)
[ http://www.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5.5/html/Technical_Notes/e4fsprogs.html ]
30/Apr/2010 @ 09:22
hugoSounds good, I did my test on a 5.4 machine and a 5.5 machine and they both had the same problems. Will investigate this a bit more in the near future.
30/Apr/2010 @ 09:25
Comment from: Bernie [Visitor] 
I use this fairly disgusting snippet in my ks.cfg:%post
(
# anaconda seems to want to do ext4 for boot, which pygrub doesn't like
uuid=$(tune2fs -l /dev/xvda1 | awk '/UUID/ {print $3}')
cd /
tar cf /var/tmp/boot.tar boot
umount /boot
mkfs.ext3 -L /boot -U $uuid /dev/xvda1
sed -i -e '/\/boot/s/ext4/ext3/' /etc/fstab
mount /boot && tar xf /var/tmp/boot.tar && rm /var/tmp/boot.tar
[...]
14/May/2010 @ 03:01
Has anyone had any problems with booting PV RHEL6/CentOS6 DomU's other than this? It hangs here.Started domain centos6 (id=183)
PCI: Fatal: No config space access function found
27/Jul/2011 @ 20:43
Hi Grant,I have dozens of CentOS-6 pv/i386 and x86_64 VMs running on CentOS-5/Xen, using all stock code included in the distro - no issues so far. there were a few stability issues under extreme and sustained load; which might even be down to i/o flakyness rather than xen itself. Havent been able to invest the time to investigate that issue
27/Jul/2011 @ 20:59
This post has 38 feedbacks awaiting moderation...