Red Hat Bugzilla – Bug 163911
kickstart probes block devices it shouldn't and halts automated install process
Last modified: 2007-11-30 17:07:19 EST
Description of problem:
Trying to deploy RHEL4 with kickstart on a SAN-cluster fails, as anaconda
detects passive paths to the SAN which return no partition table.
This is the relevant part of ks.cfg:
bootloader --location=mbr --append="rhgb quiet" --driveorder=cciss/c0d0,sda,hda
clearpart --all --drives=cciss/c0d0
part /boot --fstype "ext3" --size=100 --ondisk=cciss/c0d0
part pv.42 --size=0 --grow --ondisk=cciss/c0d0
volgroup systemzs08 --pesize=32768 pv.42
logvol / --fstype ext3 --name=root --vgname=systemzs08 --size=1024 --grow
logvol swap --fstype swap --name=swap --vgname=systemzs08 --size=1000 --grow
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.Create a kickstart file for a SAN-attached node
2.Try to kickstart-install the node
The installation halts with a prompt about /dev/sda and /dev/sdb not having a
partition table and whether they should get one.
The kickstart installation should continue regardless of the state of other
devices not used for installation.
This does not happen on FC4 with a SATA system disk on sda, and the faulty paths
beginning at sdb (here we use a Compaq Smart Array, and sda is the first faulty
It looks like FC4 does not trigger the bug because it doesn't see the SAN
devices at anaconda time.
What makes it even worse is that the non-faulty paths are now carrying a gpt
table. Now anaconda needs 4 (!) prompts during kickstart time for *each* device.
E.g. for 5 devices on the SAN the kickstart process gets interrupted 20 times.
Is there a way to communicate to anaconda not to look at any other block devices
other than those given for the installation?
There's an ignoredisk directive to ignore specific disks.
Thanks, I found the syntax as
where <driveN> is one of sda,sdb,...,hda,... etc.
But it only works if --initlabel (in clearpart) is not used. If it is used, then
initAll is set which reenables probing for all disks in DiskSet.skippedDisks.
Should I open a new bug for this one?