Re: Bootloader problems deploying from ostree repo using anaconda kickstarter (reproductible, configuration files attached)



(the adventure continues...)

Investigating more deeply the problem, I checked the initial failing
command (ostree admin instutil) fails because it does not know the
root filesystem, so it finds no deployments. From the installer
prompt, if the same command is executed, it does not fail (but I
continue having a non-bootable installation; I suspect it would be
setup in the next installation steps).

I tried to hack the code a little bit changing its call on
rpmostreepayload.py and adding the parameter --sysroot=/mnt/sysroot/
but it fails with "openat: no such file or directory", what makes me
suspect the command is called in the with chroot in the deployed tree,
which seems not to know the ostree deployment dir...

Should I fill a bug for anaconda or someone has ever used this feature
without problems?

On 24 May 2015 at 22:27, Leandro Santiago <leandrosansilva gmail com> wrote:
Just one more question: is this "ostree netinstall" currently
supported, even if in a beta state? Perhaps what I am trying to do is
just not supported, even though I think it'd be a really nice feature
(and would help me a lot! :-)).

On 24 May 2015 at 21:05, Leandro Santiago <leandrosansilva gmail com> wrote:
Well, I chrooted in the installed system and deployed manually the new tree:

# ostree admin deploy --os=centos-atomic-egym
centos-atomic-egym:centos-atomic-egym/7/x86_64/standard
--karg-append=root=LABEL=EGYM_ROOT # I changed it to use label instead
of UUID, but had the same results using UUID.
Transaction complete; bootconfig swap: yes deployment count change: 1

Now If I try to boot this system it won't work, because the grub2
module files are missing from the global /boot/grub2 directory...

Then I run manually grub2-install:
# grub2-install /dev/sda

But once I reboot the system, grub fails loading its modules, because
the module files are still missing. Although they are not present in
the global /boot/, they are in the deployed tree. There is an ostree
command which makes those copies, isn't there?

On 24 May 2015 at 20:04, Leandro Santiago <leandrosansilva gmail com> wrote:
Then I did the most obvious: commited the missing link (loader) to
directory loader.0 (no trailing slash) and the error at least has
changed :-)

Now I have the following error:
08:55:38,155 INFO program: Running... ostree admin instutil set-kargs
root=UUID=3aaf5f99-4bc2-4926-8de3-de813e99bf06
08:55:38,267 INFO program: error: Unable to find a deployment in sysroot
08:55:38,269 DEBUG program: Return code: 1

I am now trying to solve this issue...

On 23 May 2015 at 20:31, Leandro Santiago <leandrosansilva gmail com> wrote:
First of all, I found no documentation about doing what I am trying to
do, but it's in theory possible. Please let me know if I am posting to
the wrong mailing list :-)

What I am doing is basically the following: From a normal fedora
installer (I used both fedora 22 and rawhide) I am booting a machine
(for now on virtualbox) and using a remote kickstart which deploys a
remote ostree (based on centos 7, generated by rpm-ostree-toolbox). In
a ideal world it should work, but I am having some problem in the
post-install phase, installing the boot loader.

Basically the error happens then the installer tries to move the
/boot/grub.cfg file to /boot/loader/grub.cfg
(os.rename(boot_grub2_cfg, target_grub_cfg) on rpmostreepayload.py).
But as the directory /boot/loader/ does not exist, the installation
fails.

I am not using the CentOS 7 installer because it does not support
ostreesetup kickstart command, whereas Fedora does. When using Fedora
to deploy CentOS, am I trying to achieve the impossible? :-)

In order to reproduce the problem, download the attached file, extract
its contents and generate (inside it) the tree:
# rpm-ostree-toolbox treecompose -c centos-atomic-egym.ini --ostreerepo repo

Then copy the repo dir to the server machine, which in my case has the
IP 192.168.56.1 (Virtualbox host). Copy the file setup.ks too. If your
IP is different, you will need to change it in this file

Then I start a simple http server in the directory which contains both
repo/ and setup.ks, where the data will be fetched from.

# python -m SimpleHTTPServer 8080

Now trying to deploy the tree on the client... I am booting a clean
virtualbox machine (with a clean 6GB virtual hard drive, which will be
wiped anyway) which is able to reach the host on network. As I said
before, I am using fedora 22 installer, but the latest rawhide image
(boot.iso) has the same behavior.

On the boot options, I add: ks=http://192.168.56.1:8080/setup.ks

The process is non-interactive and, just after pulling and deploying
the image, the bootloader error happens.

This may not be related to ostree, but to CentOS or what I am trying
to perform is just nonsense :-) Perhaps someone in the mailing list
will be able to give me any clue :-)

Thank you in advance.

Oh, both config files and error logs generated by anaconda are in the
link: https://dl.dropboxusercontent.com/u/3550969/egym-ostree-config.tar.bz2

--
Sent from my mind



--
Sent from my mind



--
Sent from my mind



--
Sent from my mind



-- 
Sent from my mind


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]