[SOLVED!!] - Katastrophic Koala upgrade - system completely borked...

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

Rebooting will probably fix the network. When you "chroot /media/disk-1" you actually make /media/disk-1into your /. When you do that you have no /dev from which to access devices. After you run

mount -o bind /dev /media/disk-1/dev

you can run

chroot /media/disk-1

and you will have access to the devices in your chroot. The same goes for the others. I think this will allow aptitude to work better. You could also think about mounting your actual /boot at /boot in the chroot so that the linux-image will build the initrd.img into your actual /boot and you will not have to copy it over. You would do that in the chroot with standard mount -t ext2 ... If you do that be sure to umount it before exiting the chroot.

[bag]

Re: Katastrophic Koala upgrade - system completely borked...

Right then... got a bit further along the road here...

This is what I did this evening.

1. Rebooted CD-ROM and Alt-Ctrl-1 out to a command line prompt
2. Stopped KDE.
3. Bound the following filesystems (dev, proc, tmp, /dev/pts to their /media/diskX counterparts)
4. chrooted into the root filesystem on my disk
5. Did a "sudo aptitude update"
6. Did a "sudo aptitude -f install"

What a difference that made! It actually did something and resolved a few issues - before finally getting back to its old tricks of failing out on too many errors.

7. Then ran a "sudo dpkg --configure -a"

This whizzed by at the speed of light and failed extremely quickly on too many errors.

So back to the same spot really. I've also just run a "sudo aptitude dist-upgrade" for good measure. I know it's not as clever as the other commands, but still, in for a penny, in for a pound, right?

It downloaded 5MB of additional data (python stuff, binutils etc) and then unpacked it, installed and whizzed by with a load more errors... Then it pops out at the end with the now infamous "Processing was halted because there were too many errors" and lets me know that the "Current status: 411 updates [-7]"

Anyone know what the square brackets mean? Does it mean that I've just solved seven of the updates?

So that's where I am now. Still got a load of errors - still not able to update them. I'm in a bit further, and I'm sure that the whole chroot system is as it should be, with the right bits bound to my filesystems, but still, I've now gone round the loop a couple more times and I'm sure that it's not advancing any further...

I may need to fix something manually - and then see if it will advance any more.

Any ideas on where to start?

Thanks.

Bag.

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

Made some progress anyway. The place to start is the first error that is encountered. I wonder if that may be the installation of the new kernel. Since you have a separate /boot the installation of the new kernel might fail because of grub not being there.

[bag]

Re: Katastrophic Koala upgrade - system completely borked...

OK mando_hacker, I've now run the arguments to a text file for viewing.

This might take some time, but will go through the first failure points as quickly as I can.

Running the following:

dpkg --configure -a > configA.out 2>&1

returns a text file called "configA.out". When I review in nano, it says the following:

Setting up/ dbus (1.2.16-0ubunut9) ...
The system user 'messagebus' already exists. Existing.
Failed to open connection to "system" message bus: Failed to connect to socket /var/run/dbus/system_bus_socket: No such file or directory

So, having now read this error again, I think that I need to mount/bind /var so that it can access it. I've done this now - which requires exiting from the chroot, mounting the /var system and then chrooting again back into my data on the hard drive.

Point to note is that each time I come back into the chroot environment, I seem to need to run an "sudo aptitude update" to refresh the data about what needs updating. It doesn't seem to pick this up from a file on the hard drive or anything.

The reason that I think this, is that when I run a "sudo dpkg --configure -a" or a "sudo aptitude -f install" it tells me that there's nothing to do... Is that right?

Am running an update now, so that I can then try the -f install again - to see what pops out and whether the error messages will be diffferent. Will post again in a minute.

Let me know if I'm wandering off down the wrong path...

Cheers.

Bag

[GreyGeek]

Re: Katastrophic Koala upgrade - system completely borked...

Five days of slogging around trying to fix an upgrade from Jaunty to Karmic?

Is there any reason why you can't save your data to a DVD or two and then do a FRESH install from a LiveCD that has been throughly vetted?

You could have done that in an hour or two.

[bag]

Re: Katastrophic Koala upgrade - system completely borked...

GreyGeek, it's called a "learning experience'' !!!!

I know full well that I could have installed and just copied back my /home backup but I don't want to just yet. I'm enjoying the challenge of trying to understand what's gone wrong, and learn how to get my head around problems so that in the future I won't need so much help - and then I'd maybe be able to offer help back to others.

Besides, when there's so much good help on this forum, why not take the opportunity to learn from them?

I've got a little time left before I really need to get the system up and running. If I can, I want to try and fix it from the CLI. In the meantime I can check e-mail from work, and I don't need the files that are stuck on the hard drives in my machine...

Some people just like doing it the hard way!!! Normally, I'd go for the quick result, but not this time. I want to do this to see if it can be done. Cheers.

****************
So then, back to business...
Well now then, I am officially confused.

I've run a "sudo aptitude update" which did its thing, no problems.

Then I input a "sudo aptitude -f install" and it told me that there was nothing to install - although 24 packages would not be upgraded. It did also say in the first line that it was unable to resolve 'ubuntu' - which is the name of the 'machine/system' at the command line (the CD-ROM boot environment? My PC is called Kubuntu though... so I don't think it's my machine that it's talking about...)

Then I ran a "sudo dpkg --configure -a" and it just told me straight that it it could not resolve ubuntu and failed out immediately - one line of error only. That was it...

Internet connection still works though - and I know that because not only did it just download the updates, but I have now run "sudo aptitude dist-upgrade" just for good measure and it's busily downloading the 24 upgrades that wasn't going to do earlier!

Will let you know how it gets on with the upgrade when its finished.

Cheers.

Bag.

[bag]

Re: Katastrophic Koala upgrade - system completely borked...

OK, it's getting very late here - and the kids will be up in less than six hours...! So just one last thing...

Tried looking in grub again just now.

For the latest kernel option I have four lines if I try to edit it...

uuid 23fadc37-9946-43b0-87ac-de7f91470690
kernel /vmlinuz-2.6.28-16-generic root=UUID=faaa3c7c-93e3-4b24-a03f-8f95a1f616f2 ro quiet splash
initrd /initrd.img-2.6.28-16-generic
quiet

Now, not knowing a whole lot about UUID's, should the two shown above be the same What is the first UUID for?

Anyway, that's enough for now. Til tomorrow.

Bag.

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

OK mando_hacker, I've now run the arguments to a text file for viewing.

This might take some time, but will go through the first failure points as quickly as I can.

Running the following:

dpkg --configure -a > configA.out 2>&1

returns a text file called "configA.out". When I review in nano, it says the following:

Setting up/ dbus (1.2.16-0ubunut9) ...
The system user 'messagebus' already exists. Existing.
Failed to open connection to "system" message bus: Failed to connect to socket /var/run/dbus/system_bus_socket: No such file or directory

So, having now read this error again, I think that I need to mount/bind /var so that it can access it. I've done this now - which requires exiting from the chroot, mounting the /var system and then chrooting again back into my data on the hard drive.

Point to note is that each time I come back into the chroot environment, I seem to need to run an "sudo aptitude update" to refresh the data about what needs updating. It doesn't seem to pick this up from a file on the hard drive or anything.

The reason that I think this, is that when I run a "sudo dpkg --configure -a" or a "sudo aptitude -f install" it tells me that there's nothing to do... Is that right?

Am running an update now, so that I can then try the -f install again - to see what pops out and whether the error messages will be diffferent. Will post again in a minute.

Let me know if I'm wandering off down the wrong path...

Cheers.

Bag

Having been gone a while I will respond piecemeal to these posts. I think you are at the right problem, but you stopped copying it right before the most important part. The next line was probably something about "script failed and /var/lib/dpkg/info/xxx". I would like to know what that says.

The binding of /var is a reasonable thought, but it is not helpful. The reason is that when you do that the dpkg databases of which programs are installed or not are those for the CD rather than your hard disk installation. Updating that was done after binding /var will not be reflected in the on disc installation.

This dbus problem may be the holdup though. This is one reason why older unix heads are reluctant to see dbus in use, but that is another topic. I think you should umount /var and then run "service dbus start" in the chroot. The goal is to setup the proper environment inside the chroot. This may work or it may give a message about other services that have not yet been started. If so you may be able to start them in the order required.

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

So then, back to business...
Well now then, I am officially confused.

I've run a "sudo aptitude update" which did its thing, no problems.

Then I input a "sudo aptitude -f install" and it told me that there was nothing to install - although 24 packages would not be upgraded. It did also say in the first line that it was unable to resolve 'ubuntu' - which is the name of the 'machine/system' at the command line (the CD-ROM boot environment? My PC is called Kubuntu though... so I don't think it's my machine that it's talking about...)

Then I ran a "sudo dpkg --configure -a" and it just told me straight that it it could not resolve ubuntu and failed out immediately - one line of error only. That was it...

Internet connection still works though - and I know that because not only did it just download the updates, but I have now run "sudo aptitude dist-upgrade" just for good measure and it's busily downloading the 24 upgrades that wasn't going to do earlier!

Will let you know how it gets on with the upgrade when its finished.

Cheers.

Bag.

I think most of the confusion here is caused by mounting /var/ from the CD inside the chroot. In the chroot there should be a file /etc/hostname, which contains the hostname for that installation and in /etc/hosts that hostname should be associated with 127.0.0.1. That is true of the CD also.

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

OK, it's getting very late here - and the kids will be up in less than six hours...! So just one last thing...

Tried looking in grub again just now.

For the latest kernel option I have four lines if I try to edit it...

uuid 23fadc37-9946-43b0-87ac-de7f91470690
kernel /vmlinuz-2.6.28-16-generic root=UUID=faaa3c7c-93e3-4b24-a03f-8f95a1f616f2 ro quiet splash
initrd /initrd.img-2.6.28-16-generic
quiet

Now, not knowing a whole lot about UUID's, should the two shown above be the same What is the first UUID for?

Anyway, that's enough for now. Til tomorrow.

Bag.

You need to keep the grub on the CD distinct from the grub of the on disc install. To work on grub you should mount your on disc systems /boot partition on /boot in the chroot. This is not a bind operation. That partition is not mounted in your CD environment, so you will mount it using mount inside the chroot. Something like

mount -t ext3 /dev/sda1 /boot

where you use your own sdax and it might be ext2 instead of ext3. Look in /etc/fstab in the chroot.

If we get to where your system is actually upgrading itself you should not need to manually edit grub configuration. This might be essential because you will probably be moving from grub-legacy to grub2 and that will change numerous things.

PS You should check to make sure that the CD did not mount your partitions somewhere, just to be sure. The command mount will list the mounts. Do this outside the chroot.

[GreyGeek]

Re: Katastrophic Koala upgrade - system completely borked...

GreyGeek, it's called a "learning experience'' !!!!

That it is, bag!

I know full well that I could have installed and just copied back my /home backup but I don't want to just yet. I'm enjoying the challenge of trying to understand what's gone wrong, and learn how to get my head around problems so that in the future I won't need so much help - and then I'd maybe be able to offer help back to others.

You're NOW enjoying it? In your posting, "Reply #10" you stated there wasn't much joy in this slog. That's why I posted my suggestion -- just to put an end to your misery. If you are enjoying this experience it will be good for you. Ignore my suggestion!

Besides, when there's so much good help on this forum, why not take the opportunity to learn from them?

There sure is. And with Mando you are in good hands!

I've got a little time left before I really need to get the system up and running. If I can, I want to try and fix it from the CLI. In the meantime I can check e-mail from work, and I don't need the files that are stuck on the hard drives in my machine...

Some people just like doing it the hard way!!! Normally, I'd go for the quick result, but not this time. I want to do this to see if it can be done. Cheers.
...

Learning to run and admin Linux from the CLI is an admirable goal. With your determination I am sure you will reach it. Then you will be able to give back to the community as you have been given.

[bag]

Re: Katastrophic Koala upgrade - system completely borked...

Mando,

Right then, I've run the command

sudo service dbus start

It returns with the following:

sudo: unable to resolve host ubuntu
Rather than invoking init scripts through /etc/init.d, use the service(8) utility, e.g. service dbus start

Since the script you are attempting to invoke has been converted to an Upstart job, you may also use the start(8) utility, e.g. start dbus

Er, isn't that what I did?

Bag.

Will put the kids to bed and have another go later and follow up on your other comments... Cheers.

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

Just to make sure, you did this in the chroot after un-mounting the /var, right? If so it seems that the ubuntu hostname is preserved in /proc/ and/or /sys so maybe you should temporarily edit /etc/hosts in the chroot and change kubuntu to ubuntu. That should remove this problem. In fact, after editing /etc/hosts try dpkg --configure -a again and see if it might run without starting dbus.

Later you can change /etc/hosts back or change the hostname to something personal.

[bag]

Re: Katastrophic Koala upgrade - system completely borked...

Yes, I did exactly that - unmounted /var, then tried to start the dbus system.

Have now changed /etc/hosts to ubuntu (was previously kubuntu). And have run the command

sudo service dbus start

again... And it returns exactly the same message. Although, I have just noticed that although it says this - it does not actually seem to be an error message... It has a response that it posts up - but it does not say that it didn't start...

How can I quickly check the actual status - to see if it is running or not?

Thanks,

Bag

[flarson]

Re: Katastrophic Koala upgrade - system completely borked...

ps axf | grep dbus

should show a process running. You can also try

dbus-daemon --introspect

although I don't know the implications of this. It should output an XML configuration, but whether it does this through the running daemon I am not sure. Maybe someone else here can inform us about that.