Created attachment 2177 [details]
report.bug.zip

It seems / partition is full

# df

http://dl.dropbox.com/u/4147101/DSC01037.JPG

After removing /usr/share/doc to clear some space and retrying to upgrade it gets to Package Group Selection and then..

A fatal error occurred: unable to open rpmdb (9).

If it matters, this was an upgrade from mga1 to mga2 prerelease RC with the latest DVD x86_64 (v5) dumped onto usb stick. 

I had installed everything I could think of into mga1 to test the upgrade and there was something like 2GB free space on /.

Clicking OK to the rpmdb (9) error gives another error..

An error occurred
Can't call method "traverse_tag" on an undefined value.

Your rpmdb errors show the fs is full...
Due to this rpmdb might have been corrupted hence the error

It wasn't full before the error, there was approximately 2GB free on / (comment 5)

Yes but it got filled.
Probably b/c of the unique transaction (old packages get removed only at end).
For the record, could you attach your /etc/urpmi/urpmi.cfg?

If you are able to reproduce, please stop the install, reboot and send me the bug5746.tar.xz files resulting from:

urpmi --auto-select --bug bug5746
tar cfa bug5746{.tar.xz,}

Or do it from the installer console (just run "chroot /mnt" first)

Isn't the installer supposed to check there is enough space before installing? 

I remember failing upgrade from mandriva due to not enough space (bug 2383 & bug 1342) but from what I have heard this relates to all packages being downloaded at once to /var/tmp before they are installed and so filling up the disk.

The installation was unbootable, only into failsafe and then couldn't mount any usb storage so has now been overwritten.

s/failsafe/old kernel/

Old kernel was unusable to boot it properly but I was able to log in on tty2. Grub wasn't updated. I gave up trying to make it boot from the new kernel and moved on.

It does check.
But when URPM fallback to one big transaction, then the needed temporary space is quite much bigger (and hard to count)

This has occurred twice now out of 3 or 4 upgrades.

I'll attach two logs, one is from chroot in the installer which left mainly 0 byte files and the second is after reboot.

After the reboot it had removed mga1 medias and left only the mga2 dvd so it should be valid.

Created attachment 2194 [details]
after-reboot-bug5746.tar.xz from after the reboot

This is the file from after the reboot with much more information in it.

The chroot tar.xz was a zero byte file and it won't let me upload it. It should contain one file which wasn't but it was only a 67 byte urpmi.cfg which showed 'core media'.

$ cat bug5746/urpmi.cfg
{
}

core\ media cdrom://x86_64/media/core {
  key-ids: 80420f66
}

I created a snapshot in virtualbox, so can return to it if you need any further details, and completed the upgrade process. With 10 seconds left it sat for a long long time with nothing changing in the installer, presumably whilst it was installing all the packages. 

It did complete though and moved on to creating the bootloader, which failed. I don't know if it is related in any way but that is bug 5774.

After rebooting, it found updates with only the DVD as media.

It seemed to have circular dependencies as it asked to choose between default-windowmaker-desktop and mageia-windowmaker-desktop, then gave a choice of python-setuptools, python-virtualenv or python3-setuptools.
Choosing python-setuptools just asks again to make the same choice and then if chosen for a 2nd time says python-zope-interface cannot be selected.

Again, I'm not sure if that is related in any way.

I tried to reproduce this on a mag1 64bit server box.It had:
hda 152GB
/          hda1  60gb
swp        hda6  3gb
/home      hda6  73gb

hdb  74gb
/var  /www hdb1  28gb
/var/ftp   hdb5  46gb

I installed an extra 3.1gb of assorted software, which left about 56gb free and
did updates before rebooting and starting the upgrade.
There were error about some missing packages, about 30. I quickly scanned the
file names and they seemed like they were from most of the extra stuff I had
added. After logging in the anakonda thing moved address books and vcards.
Konsole, konversation and network manager all started up, they were running
when I shut down the machine before the upgrade.
After leaving the box to settle down it then notified me of some updates and
had problems. I'm investicating the update problem at the moment.

CC: (none) => led43john

This doesn't seem to be affected by the size of the / partition. It happened to me last time with 15GB free space there. The rpmdb (9) previously was just because the 'one big transaction' had filled the partition.

I would imagine that would happen when this occurs if more than around 50% of the / partition is used, which is likely to be most of the time.

Lack of space needs some more time so that manage it can be improved, rather for Mageia 3

Priority: release_blocker => High
Target Milestone: --- => Mageia 3

Hi,

This bug was filed against cauldron, but we do not have cauldron at the moment.

Please report whether this bug is still valid for Mageia 2.

Thanks :)

Cheers,
marja

Keywords: (none) => NEEDINFO

This bug report is about using mgaapplet-upgrade-helper to upgrade,
so the bug report should be kept as a cauldron bug report.

Keywords: NEEDINFO => (none)
CC: (none) => davidwhodgins

*** Bug 8901 has been marked as a duplicate of this bug. ***

CC: (none) => alainderaedt

@MrsB: is this not needing whiteboard: 5beta1 ? Thanks

Why? the last comments date back from 2012!

I've not seen this recently, but upgrade testing has been limited so far with so much other stuff going on.

Let's close it and reopen later if it still occurs.

Status: NEW => RESOLVED
Resolution: (none) => OLD

(In reply to Thierry Vignaud from comment #25)
> Why? the last comments date back from 2012!

Because I was testing whether adding some (but not all) old upgrade bugs to the tracker would give more chance of someone confirming it got fixed or still exists, then just pinging in old upgrade bugs.

CC: (none) => marja11