View Issue Details
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0017575||CentOS-8||anaconda||public||2020-07-08 05:41||2020-09-16 15:52|
|Target Version||Fixed in Version|
|Summary||0017575: New ISO images cannot be installed (with kickstart file?)|
|Description||In Cockpit team we are using centos-8-stream isos to create our testing images. Recently isos were replaced with new ones but they fail to be installed by anaconda.|
I don't think it is anaconda bug, but had no idea what to report against and since it fails in anaconda I thought it is a good enough starting point.
|Steps To Reproduce||```|
$ wget http://mirrors.mit.edu/centos/8-stream/isos/x86_64/CentOS-Stream-8-x86_64-20200629-dvd1.iso -O centos-8-stream.iso
$ qemu-img create -f qcow2 /var/tmp/centos-8-stream-2.qcow2 12G
$ virt-install --connect=qemu:///session --initrd-inject=fedora.ks '--extra-args=ks=file:/fedora.ks console=ttyS0,115200' --name=centos-8-stream-2.qcow2-builder --disk path=/var/tmp/centos-8-stream-2.qcow2,format=qcow2 --ram 4096 --vcpus=1 --os-type linux --os-variant fedora21 --location=centos-8-stream.iso --nographics --noreboot
$ cat fedora.ks
network --bootproto dhcp
firewall --enabled --ssh
timezone --utc America/New_York
bootloader --location=mbr --append="console=ttyS0,115200 rd_NO_PLYMOUTH"
clearpart --all --initlabel
chmod 700 /root/.ssh
echo "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDUOtNJdBEXyKxBB898rdT54ULjMGuO6v4jLXmRsdRhR5Id/lKNc9hsdioPWUePgYlqML2iSV72vKQoVhkyYkpcsjr3zvBny9+5xej3+TBLoEMAm2hmllKPmxYJDU8jQJ7wJuRrOVOnk0iSNF+FcY/yaQ0owSF02Nphx47j2KWc0IjGGlt4fl0fmHJuZBA2afN/4IYIIsEWZziDewVtaEjWV3InMRLllfdqGMllhFR+ed2hQz9PN2QcapmEvUR4UCy/mJXrke5htyFyHi8ECfyMMyYeHwbWLFQIve4CWix9qtksvKjcetnxT+WWrutdr3c9cfIj/c0v/Zg/c4zETxtp cockpit-test" > /root/.ssh/authorized_keys
chmod 600 /root/.ssh/authorized_keys
mkdir -p /etc/sysconfig/network-scripts
ONBOOT="yes"" > /etc/sysconfig/network-scripts/ifcfg-eth0
ONBOOT="no"" > /etc/sysconfig/network-scripts/ifcfg-eth1
sed -i 's/GRUB_TIMEOUT.*/GRUB_TIMEOUT=0/; /GRUB_CMDLINE_LINUX=/ s/"$/ net.ifnames=0 biosdevname=0"/' /etc/default/grub
grub2-mkconfig -o /boot/grub2/grub.cfg
systemctl is-enabled sshd.socket || systemctl is-enabled sshd.service || systemctl enable sshd.socket
echo "https://registry.npmjs.org/" > /root/npm-registry
It is stuck at:
1) [x] Language settings 2) [x] Time settings
(English (United States)) (America/New_York timezone)
3) [!] Installation source 4) [!] Software selection
(Error setting up software (Error checking software
5) [x] Installation Destination 6) [x] Kdump
(Automatic partitioning (Kdump is enabled)
7) [x] Network configuration 8) [ ] User creation
(Wired (ens2) connected) (No user will be created)
Please make a selection from the above ['b' to begin installation, 'q' to quit,
'r' to refresh]: b
Please complete all spokes before continuing
Steps 3 and 4 are not completed. It worked with previous  isos. It also works with other OSes we use, such as fedora, rhel... (although we don't use iso files, but the installation process is the same).
|Additional Information||When I took the ISO and just ran it with gnome boxes it installed correctly. It seems that the problem might be related to using kickstart and unattended installation?|
|Tags||No tags attached.|
Happens with me, too. Drove me nuts yesterday. I can install manually with the CentOS-8.2.2004-dvd1.iso, but not with kickstart. It's like anaconda isn't finding the kickstart file. With CentOS 8.1 and earlier, it did just fine.
In my case, I'm using packer to build VMs in vSphere, and I can get onto the VM's consoles while packer is booting them up. I can see them go to the graphical installer and just sit. If I interceded and tab to add the inst.text kernel command line option, it'll start the text installer, but it's clear from the settings in that screen that it hasn't seen the kickstart file.
I'd be happy to do any troubleshooting you like.
A collegue reminded me about labeling the ISO Volume as OEMDEV and once I did that, The 8.2 anaconda is now reading my kickstart file and it all works now.
"mkisofs -V OEMDEV -o kickstart8.iso ./ks.cfg " produced an ISO with the ks.cfg in it that anaconda find.
glad to hear that you were able to resolve it. However I am not sure if I understand your solution. Could you please elaborate a bit more?
`mkisofs -V OEMDEV -o kickstart8.iso ./ks.cfg`
What is the content of `ks.cfg`? Does the official iso file come into play somehow?
@Mike: So you inject the kickstart through an iso file? We just use virt-install's documented way through `--initrd-inject` to get the file into the installer env, and `--extra-args="ks=file://path/to/ks`. This has worked for years, and still works fine for anything except the very latest CentOS 8 stream, so it's definitively a regression.
virt-install doesn't seem to have a straightforward way to boot with two ISOs. Do you use `--disks` multiple times?
|We're using Packer to create VMs as templates in VMware. As far as I can tell, it's simply attaching two DVD/CD-ROM drives. The first has the first ISO I list, the second has the second ISO I list, which has the ks.cfg file in it's root.|
|This seems fixed on CentOS-Stream-8-x86_64-20200910-dvd1.iso , so please close. Thanks!|
|2020-07-08 05:41||mmarusak||New Issue|
|2020-07-29 20:43||MikeDiehn||Note Added: 0037447|
|2020-07-30 15:33||MikeDiehn||Note Added: 0037451|
|2020-08-04 07:52||mmarusak||Note Added: 0037484|
|2020-08-06 08:04||Martin.Pitt||Note Added: 0037504|
|2020-08-10 15:42||MikeDiehn||Note Added: 0037515|
|2020-09-16 15:52||Martin.Pitt||Note Added: 0037704|