Comments on: Fedora 11 fail because of Anaconda :(

By: Jhosep

Jhosep — Tue, 14 Jul 2009 08:32:20 +0000

Anaconda is bad…….
I don´t know because Fedora used this…..
change that … or better this……

By: kf2

kf2 — Fri, 29 May 2009 22:19:45 +0000

I imagine 6 months isn’t time enough for something like a total rewrite of a backend to be going into production and throwing more developers on a problem doesn’t necessarily help.

But then, fedora has been continually slipping into many regressions ever since F9 and no one seems to mind… or, likely, those that do, have already left.

By: Adam Williamson

Adam Williamson — Fri, 29 May 2009 18:23:43 +0000

Not…really. You’re not exactly understanding the process.

First, everyone was entirely aware there will be regressions in Anaconda in F11. You say “The inner working of anaconda previous to Fedora 11 was in top shape”, but it wasn’t. The storage code was old and crufty and holding a lot of things back. So for F11 the Anaconda team bit the bullet and entirely rewrote it.

That’s why a lot of stuff broke – it’s inevitable with a rewrite, but we really needed to do the rewrite regardless.

The second slip was for one specific bug in Anaconda, but it’s given us enough time that it’s worth re-evaluating the full list of Anaconda bugs to see if any others are sufficiently severe to be worth fixing this week. This was not in any way done in response to your post, sorry – it’s just part of the ongoing blocker review process being done by QA and RelEng in co-operation with the development teams involved (in this case, the Anaconda team).

Your post didn’t tell anyone involved anything they didn’t already know. :)

By: valent

valent — Fri, 29 May 2009 12:40:36 +0000

This is the reason why F11 is slipping;

Bill Nottingham:

Not knowing if this had been formally done or not, I figured I’d look.

I looked at all open rawhide anaconda bugs opened on March 1 or later.
(126 bugs). The following list of 28 bugs are all crashers or things
that render the system non-bootable.

Some of these may be fixed already; some may not be fixable, some may
have workarounds we can put on the common bug page. But if there are
any other open bugs like 500808, they’d be in this list; if someone
wants to do a sanity review to see if any of these are worth holding
out for, go for it.

https://bugzilla.redhat.com/show_bug.cgi?id=489991
Anaconda crashes when disks with no partitions contain a filesystem

https://bugzilla.redhat.com/show_bug.cgi?id=491754
DeviceError: cannot commit to disk sda after 5 attempts

https://bugzilla.redhat.com/show_bug.cgi?id=493058
Custom partitioning creation/edit causes traceback

https://bugzilla.redhat.com/show_bug.cgi?id=493699
Traceback while installing with lvm/luks

https://bugzilla.redhat.com/show_bug.cgi?id=494150
anaconda exception when formatting drives

https://bugzilla.redhat.com/show_bug.cgi?id=495139
Anaconda traceback when installing with another LVM already existing

https://bugzilla.redhat.com/show_bug.cgi?id=495356
Unhandled exception at the end of partitioning

https://bugzilla.redhat.com/show_bug.cgi?id=495433
Anaconda crash: Error opening /dev/sdb: No medium found

https://bugzilla.redhat.com/show_bug.cgi?id=496440
RAID1 volume create failed

https://bugzilla.redhat.com/show_bug.cgi?id=497183
Installation fails with partition scheme

https://bugzilla.redhat.com/show_bug.cgi?id=497293
allocatePartitions extended partition handling needs work

https://bugzilla.redhat.com/show_bug.cgi?id=498240
Anaconda exception on upgrading from F10

https://bugzilla.redhat.com/show_bug.cgi?id=498350
Anaconda crashes in the partitioner screen

https://bugzilla.redhat.com/show_bug.cgi?id=498408
Anaconda crash during install: MDRaidError: mddeactivate failed for /dev/md1

https://bugzilla.redhat.com/show_bug.cgi?id=498486
MDRaidError: mdactivate failed for /dev/md2

https://bugzilla.redhat.com/show_bug.cgi?id=498553
FSError: filesystem has not been created

https://bugzilla.redhat.com/show_bug.cgi?id=498751
Failure at the start of Live install – device is already mapped

https://bugzilla.redhat.com/show_bug.cgi?id=495156
FSError: mount failed: (2, None)

https://bugzilla.redhat.com/show_bug.cgi?id=498820
ValueError: Cannot remove extended partition sda4. Logical partitions present.

https://bugzilla.redhat.com/show_bug.cgi?id=499062
SystemError: (2, None)

https://bugzilla.redhat.com/show_bug.cgi?id=499733
Anaconda does not see Individual SATA11 drives — linux nodmraid is ineffective

https://bugzilla.redhat.com/show_bug.cgi?id=500132
swap format error – SwapSpaceError: format has not been created

https://bugzilla.redhat.com/show_bug.cgi?id=500995
TypeError: float() argument must be a string or a number

https://bugzilla.redhat.com/show_bug.cgi?id=501297
AttributeError: ‘DeviceFormat’ object has no attribute ‘mount’

https://bugzilla.redhat.com/show_bug.cgi?id=501340
FormatSetupError: format has not been created

https://bugzilla.redhat.com/show_bug.cgi?id=501584
Initial initramfs created during installation has no keymap support

https://bugzilla.redhat.com/show_bug.cgi?id=502092
RuntimeError: Returning partitions to state prior to edit failed

https://bugzilla.redhat.com/show_bug.cgi?id=502663
backtrace when setting up md raid.

By: valent

valent — Sun, 10 May 2009 18:27:47 +0000

Sorry for not replying to your questions and suggestions, but I was out of town on Croatian Linux conference. I hope to get to these issues soon and report back here.

By: Paul W. Frields

Paul W. Frields — Mon, 04 May 2009 17:04:00 +0000

I did an installation of Rawhide to my laptop last weekend. That system has a hard disk with encrypted LVM and many pre-existing partitions. And I had no such problems like the ones you describe here. Maybe you’re using something out of date? Anaconda is not the same now as it was at F11 Beta stage, and in fact it’s now even had some additional fixes since the Preview was released.

Can you point me to the bugs you filed showing your problems, so we can address them?

By: Bucky

Bucky — Mon, 04 May 2009 16:06:15 +0000

I haven’t tried it in a couple iterations of Fedora, so this may long since have
been fixed, but I gave up trying to do package customization in Anaconda. For
me, it has always ended up in a crashed installer.

By: Jon Pritchard

Jon Pritchard — Mon, 04 May 2009 12:10:59 +0000

I remember when I finally wanted to just make manual partitions in the installer without using a gparted livecd, since I’d grown tired of LVM not easily allowing me to upgrade Fedora keeping my home directory. This is not exactly easy and requires a lot of knowledge and pre-thought. You cannot choose to make a certain extended partition and then put primary partitions in it, it’ll just make an extended partition after the limit is reached.

Another time when I was installing from livecd it just presumably took forever to repartition my NTFS drive (Which is a good feature) but it didn’t give any feedback at all, it appeared as if it had crashed, after leaving it for 2 hours it had finally done its job and moved on to the next stage of installation. A less experienced/patient user could’ve ruined the system by power cycling.

Finally another gripe of mine is that Fedora to my knowledge doesn’t detect other Linux installations and create a GRUB entry for them to make them bootable.

When I used Fedora 5, for the first time I was impressed by the look of the installer, now it looks too old and unfriendly. It’s one of the most important features, if you get a crash half way through repartitioning drives etc. you could lose all your data and be left in a no boot situation.

In shot, it needs a lot of work. I was thinking that the storagerewrite would make it more reliable but what you’re saying seems to shows the opposite effect.

By: nicu

nicu — Mon, 04 May 2009 08:41:32 +0000

I had a similar crash when installing F11 Beta on my desktop: it crashed after installing all the packages. But I was kind of lucky: after a couple of minute of panic I restarted the computer, F11 started, the boot process took an incredible long time until stopped. Rebooted again and it worked.
Now I must gather enough courage and try installing the Preview Release on my laptop… Wish me luck :D

By: Smooth Criminal

Smooth Criminal — Mon, 04 May 2009 05:55:51 +0000

Yeah, I pretty much had to partition whole disk fron CLI and use ext3 because ext4 was not present (or at least no mkfs.ext4). Then i went with custom partitioning sheme without format option in GUI (just assign partitions to their mount points).
As it goes for graphical design of anaconda: it tottaly KICKS ASS. I love it! Is is simple yet feels so distinct.
And for the problems – it all comes to tinkery/hackish nature of person: I did it ;)