On Fri, 2013-12-27 at 08:42 -0500, Robert Moskowitz wrote:
Thanks for responding. Trimed down to things to reply to...
On 12/27/2013 02:59 AM, Adam Williamson wrote:
> On Wed, 2013-12-25 at 18:11 -0500, Robert Moskowitz wrote:
>> If I selected my local repo, Updates became not an option. Regardless if
>> I used the DVD install (i386 and x86_64) or the netinstal (only tried
>> x86_64), consistantly this became greyed out. I could not provide the
>> URL for where I have my local updates repo. It did not matter if I added
>> repo=url to the boot line, or did it in the GUI. The moment I selected
>> my own http URL, I lost updates.
> With the interactive install, I believe anaconda doesn't expect you to
> pass multiple repos; it's expecting either a (single) actual yum package
> repository, or a mirror tree with a .treeinfo file specifying the
> location of the standard repo set.
First is there a way to specify the updates repo on the boot line. There
was (I believe) back on F17. Or maybe I am just suffering from a senior
I don't know, it's not something I do myself and I haven't had occasion
to investigate it. The parameters themselves haven't really changed much
since F17 (they changed a lot since F15 or F16, whenever we did
'noloader') but anaconda's behaviour wrt repos did change a bit with
Where are there instructions for making a .treeinfo file and can I
specify it in the repo=url boot parameter?
Again, sorry, I don't know, not anything I've needed to know how to do.
It's part of making a Fedora mirror tree, releng would know if it's
something you can do. You don't specify a treeinfo file exactly, but if
the directory you point to with repo= contains one, anaconda will use
Now I go and look at an actual treeinfo file it may not even be the
I just know that when you point anaconda at an actual properly-laid out
Fedora mirror tree, it finds both the 'stable' and 'update' repos.
have to go poking at the code to remember preciselyt how.
> With a kickstart you can specify multiple separate repos, I
think. But I haven't really poked into this
> behaviour much since newUI.
My 'practice' is to first get an install working, then use the
anaconda.cfg to build a kickstart file. So here I am, not getting to 1st
Well aside from the other bugs, the obvious thing would be to do your
base install without multiple repos and then add them once you get to
working on your kickstart.
I almost discourage you from poking around. I feel it is a real
dumb down, and very saddening. Particularly that I can not customize
what apps and groups to install or not. Just a large general catagory.
There's generally a reason for dumbing things down, and here the reason
is simple: complexity in repos and package selection is a heavy
maintenance burden for a small reward since the majority of people
install from the media or the official mirrors and are perfectly okay
with customizing package set post-install or in a kickstart. There
really aren't very many cases where it's significantly useful to be able
to customize the package set on a per-package basis, interactively,
during installation, and to back that fairly fringe feature, anaconda
devs had to maintain an entire packaging GUI inside the installer, which
was a lot of work (and a potential source of lots of bugs). As I said I
don't know the full details of how multiple repos work/don't work
interactively vs. cmdline vs. kickstart, but any 'simplification' in
this area likely has a similar explanation.
>> Sometimes when I selected LVM for the partitioning, the LVM
>> name would be 'fedora_19'. I left it like that in this install that
>> finally worked and here is what df has to say for itself:
>> $ df -h
>> Filesystem Size Used Avail Use% Mounted on
>> /dev/mapper/fedora_19-root 29G 4.8G 23G 18% /
>> devtmpfs 1.3G 0 1.3G 0% /dev
>> tmpfs 1.3G 164K 1.3G 1% /dev/shm
>> tmpfs 1.3G 1016K 1.3G 1% /run
>> tmpfs 1.3G 0 1.3G 0% /sys/fs/cgroup
>> tmpfs 1.3G 44K 1.3G 1% /tmp
>> /dev/sda1 477M 96M 352M 22% /boot
>> /dev/mapper/fedora_19-home 257G 32G 212G 14% /home
>> Some times it used the host name. Depended on what steps I took to get
>> to setting up the LVM partition.
> It's probably re-using the existing VG rather than blowing it away and
> creating a new one, the '19' rather suggests that.
On a new, empty, SSD drive from Crucial or a old drive that had f17 on
it? Never installed f19 here. I think the disk druid developers need
to search their code for some label they did not change.
If there is one, it's not particularly obvious...
[adamw@adam anaconda (f20-branch %)]$ grep -R fedora_19 *
[adamw@adam anaconda (f20-branch %)]$ grep -R vgname *
pyanaconda/kickstart.py: vgname = ksdata.onPart.get(self.vgname,
pyanaconda/kickstart.py: vg = devicetree.getDeviceByName(vgname)
KickstartValueError(formatErrorMsg(self.lineno, msg="No volume group
exists with the name \"%s\". Specify volume groups before logical
volumes." % self.vgname))
pyanaconda/kickstart.py: if not self.vgname:
pyanaconda/kickstart.py: dev =
KickstartValueError(formatErrorMsg(self.lineno, msg="No preexisting VG
with the name \"%s\" was found." % self.vgname))
pyanaconda/kickstart.py: elif self.vgname in (vg.name for vg in
KickstartValueError(formatErrorMsg(self.lineno, msg="The volume group
name \"%s\" is already in use." % self.vgname))
pyanaconda/kickstart.py: ksdata.onPart[self.vgname] =
[adamw@adam anaconda (f20-branch %)]$ grep -R request.name *
pyanaconda/kickstart.py: ksdata.onPart[self.vgname] =
probably in there somewhere, but meh. Again, doesn't seem
burning-down-the-house serious, but worth filing a bug on, sure.
>> In summary of what is important:
>> Why have my installs to the SSD failed (writing bootloader).
> It is absolutely impossible to tell without at least program.log. It's
> kind of annoying that all cases of bootloader install failing on UEFI
> install keep getting written off as dupes of 1006304, because when
> they're marked as dupes we don't get logs, and there's just no way to
> know what's going on. I've posted a couple of comments asking if
> anaconda and libreport devs can figure that out, but nothing doing yet.
> If you have a copy of program.log (and ideally all the other logs...)
> from one of those failures, we could try and figure it out.
Instructions on generating (or capturing) the logs?
They're in /tmp while the installation is running, you can scp or fpaste
or copy-to-a-USB-stick them out from there.
I would be glad to
try. Though I am thinking of building a netinstal USB stick and seeing
if there are timing problems with the USB DVD drive. The observed
symptom is the drive starts up (it had been off during the post install
process) and a spinning wait object is going on the screen. After some
time (minute or so?) I get the error.
It's worth trying, but it could well just be an issue with your UEFI
firmware in some way. Again, impossible to tell without the logs,
unfortunately (all the anaconda error message tells us is that EFI
bootloader configuration failed; it has zero indication of why, that's
always in the output from efibootmgr in program.log).
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net