On 11/30/20 11:34 PM, Christian Borntraeger wrote:
On 30.11.20 21:10, Mark Post wrote:
On 11/30/20 1:16 PM, Alan Haff wrote:
I have a number of Linux guests running under z/VM. I'd like to move some of
them to a KVM host running in another LPAR. Naively I tried DDRing a guest's
disks to new volumes and IPLing off of the new volumes. No luck, the virtio
modules aren't available in the guest so it can't see any DASD.
Has someone put together a recipe for moving a guest from z/VM to KVM? Or am I
heading down a dead end?
I'm not aware of one, but I wouldn't call it a dead end, either. As a
first step, I would rebuild the initrd with whatever parameters are
needed to include the virtio drivers, then try again. I have to think
that /etc/fstab entries would need to be looked at as well.
Look at the virtualization cookbook 5 from the redbooks.
http://www.redbooks.ibm.com/abstracts/sg248463.html?Open
chapter 8.7 contains information how to configure a system as such that it can
be moved forth and back between LPAR and KVM. That should then also be possible
with z/VM. If done right that should allow you to move the guest forth and back
between KVM and z/VM.
Basic idea:
make sure the initrd has classic (DASD) drivers and the virtio-blk driver and
then use driver independent names for the partitions.
CC Viktor, do we still have our presentation flying around somewhere?
I co-authored Chapter 8 and did much of the testing. I agree with
Christian that z/VM -> KVM is possible.
Chapter 8 starts with a SLES KVM VM running on a SLES KVM host running
in an LPAR. The SLES OS is installed on FCP SCSI disks. The KVM VM and
host are shutdown and the SLES OS on the SCSI disk is IPL'd in the same
LPAR. Using DDR to duplicate the z/VM guest to the new DASDs on your
KVM host is a good first step. The good thing about this approach is
that you can continue to run the z/VM guest as you work through the
hurdles to get the KVM host to boot.
I suggest keeping the device ids of the OS disks defined in the KVM
guest the same as it is in z/VM. I encourage you to review the
virt-install command in the shell script in Appendix B as an example how
to define the KVM VM.
I would define the NIC with the same device id but add that to a NAT'd
KVM network. This will allow the network of the new KVM VM to start but
will not step all over the IP address being used while the z/VM guest is
running at the same time.
I also agree with Mark that it might be necessary to rebuild the initrd
so that virtio drivers are available during IPL.
Regards,
Mike
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit
http://www2.marist.edu/htbin/wlvindex?LINUX-390
