On Fri, 2013-07-12 at 02:58 -0400, Felix Miata wrote:
On 2013-07-11 22:07 (GMT-0600) Chris Murphy composed:
> When you click + a prior selection is dropped, and a new mount point is
> created from free space based on the parameters: mount point (/boot, /,
> swap, etc) and size. So whatever you selected before clicking + isn't
> relevant.
So on first arriving on the
http://fm.no-ip.com/SS/Fedora/anacondaF19manpart01.png screen, what needs
doing in what order to not see the red failed to add device message appear?
Do I ever need to click the "+" button if the only partition to configure is
/?
If you want to create a new volume to back /, yes. If you don't, no.
Why is it the "configure selected mount point" button is
always grayed
out, including right after having hit plus button and filled out the mini
window with mount point and "desired size"?
Probably because there would be nothing you could actually do in it.
It's a fairly minimal screen: right now, IIRC, all it lets you do is
decide which disk a to-be-created volume will reside on. If you only
selected one target disk, then this is obviously moot; it's also moot
for existing partitions, as they clearly can't be moved from disk to
disk.
> I recall an unreproducible logic bug regarding size, where I
definitely
> had enough free space like in your screen shot, yet anaconda would fail to
> add new device with the reason that there wasn't enough free space. And
> the only way I found out of this was to restart anaconda.
Did I say I hate newUI? I HATE IT I HATE IT I HATE IT. I've done hundreds,
maybe thousands, of Linux distro installations sprawled over more than a
decade. None have been anywhere near as difficult as F18 or F19.
Well, look, you're a reasonably smart person. The concept of how newUI
custom partitioning works is not difficult, and extensively explained in
the wiki, the page's own Help screen, and the Installation Guide; but
you seem to persist in confusing the 'create a new storage volume'
action and the 'modify a defined storage volume' action. As long as you
do that, you're going to be having trouble.
(Also, frankly, the UI is not optimized for crazy layouts with 20+
partitions on a disk, because they are the unusual case; the UI is
optimized for the more common cases, which is what you should always
optimize a UI for.)
After I read your reply I remembered this lack of space problem had
happened
some time before F19 got to RC1, and that formatting the target before
starting Anaconda seemed to be a workaround. So far tonight that isn't working.
Ever since I started this thread several hours ago, except for about 30
minutes of break time, I've been thwarted from getting an installation
started in too many ways to remember. Working backwards, I'm now making less
progress trying than earlier, repeatedly stopped by
https://bugzilla.redhat.com/show_bug.cgi?id=974467 and facilitating me
forgetting whatever it was I wanted to try differently next.
Navigating through manual partitioning (which I don't need, as all I need to
do is to set the / mount point) requires unintuitive procedural logic. Input
fields filled in or buttons clicked out of secret order are treated as though
left blank or unclicked.
Are you hitting the 'Update Settings' button after you change something?
That's what it's for. Apparently, there are significant problems to
trying to have changes simply auto-apply when entered.
When I do manage to get through it, I'm always
stopped by an Anaconda claiming failed to add device, lack of available
space, primary partition required (on $SUBJECT attempts), or others I just
can't remember after too
many tries to count.
It's going to be pretty hard to diagnose any of these without more
precise information.
Usually the only thing you absolutely need is a partition assigned the /
mount point. But if you're doing a UEFI-native install, you also need a
partition of the EFI System Partition type, mounted at /boot/efi . If
you're doing a BIOS-native install to a GPT-labelled disk (which may be
the case, I guess, since you have eleventy billion partitions), you need
a BIOS boot partition to be present on the target disk.
Trying to exit and start over takes what seems like forever each
time.
Anaconda's GUI closes down soon enough, but then nothing happens for much too
long waiting on NM to time out doing some useless thing it should have no
need to do to effect a shutdown and reboot that I would have used the reset
button for if the machine had one.
That's a known issue:
https://bugzilla.redhat.com/show_bug.cgi?id=974383 . The fact that your
machine has no reset button is, I suggest, a known issue in your
hardware. =)
Mind you these are all started with
complete network configuration on cmdline, so there should be nothing network
related to "manage" during the Anaconda installation or abort processes.
That conclusion does not really follow from that premise. It's not an
'anaconda installation or abort process', it is just systemd starting up
and shutting down the system. NetworkManager is used for...network
management...in the installer environment. The fact that it delays is a
bug, obviously, as linked above.
Is there a Rawhide/Anaconda kernel/initrd/squashfs.img installation
set
somewhere that I can try instead of F19's?
That would not be a good idea. If the question is "I'm having installer
problems, what might help out?", the answer is rarely "Use an unstable
and almost entirely untested version of the installer!"
(So far as I know no-one outside the anaconda team has tested the F20
installer in a non-live configuration, yet. In a live configuration it
appears to contain a 100% showstopper bug at present.)
Telling excerpts from anaconda.log:
06:25:38,953 DEBUG anaconda: disk free: 59.06 GB fs free: 348.67 GB sw
needs: 766.56 MB auto swap: 3.98 GB
06:25:45,481 INFO anaconda: spoke is ready:
<pyanaconda.ui.gui.spokes.storage.StorageSpoke object at 0x7ff4e23dead0>
06:26:33,311 DEBUG anaconda: requested size = 4.8 GB ; available space =
59.06 GB
06:26:33,444 ERR anaconda: factoryDevice failed: not enough free space on disks
06:32:32,721 DEBUG anaconda: info bar clicked: not enough free space on disks
((<SpokeWindow object at 0x7ff4e22dfa50 (AnacondaSpokeWindow at 0x388e620)>,))
Whether or what bug to find or file to attach them to is up for suggestions.
That sounds almost like the 'native UEFI install to an msdos-labelled
disk' bug, but I doubt that's quite what you're hitting. Still be hard
to tell for sure without full anaconda.log and storage.log though. But
honestly, given your travails with grokking the interface, the best
thing might be a screencast...
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net