installing lucid over existing software raid 1 does not work

Bug #560152 reported by stop
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
partman-base (Ubuntu)
New
Undecided
Unassigned
Nominated for Lucid by stop

Bug Description

Binary package hint: partman-base

I have a machine with 3 hard-disks, /dev/sda for / and swap and /dev/sdb and /dev/sdc for a RAID 1 /home.
This works fine in karmic, an upgrade from Karmic to Lucid (update-manager -d) also works fine now.
However it seems feasible that you would like to overwrite / with a fresh ubuntu install and leaving the /home RAID 1 intact. One would probably use the alternate cd installer for this.
During tests, it does not seem to be possible.. I have tried leaving as much as is during the install, erasing all data from /dev/sda, recreating the RAID array but after finishing partitioning I always get:
--------------------------------------------
Failed to remove conflicting files
The installer needs to remove operating system files from the install target, but was unable to do so, The install cannot continue
--------------------------------------------

See screenshot for partitioning scheme...

The same happened for a server install (with ubuntu-10.04-beta2-server-amd64.iso). It had two hard disks with 3 raid arrays: one for /, one for swap and one for /home. When booting the iso and trying to keep the data on the /home raid array while overwriting the / array it got the same error message and installation failed.

See additional screenshots for server partitioning scheme and error message...

Revision history for this message
stop (whoopwhoop) wrote :
description: updated
Revision history for this message
stop (whoopwhoop) wrote :
summary: - installing lucid over / but leaving /home raid 1 intact does not work
+ installing lucid over existing software raid 1 does not work
description: updated
Revision history for this message
stop (whoopwhoop) wrote :
stop (whoopwhoop)
tags: added: lucid partman raid
Revision history for this message
stop (whoopwhoop) wrote :

This is still not working for me in the release candidate of lucid...

Revision history for this message
NickJ (nickd-jones) wrote :

Me neither, in fact choosing "Keep existing data" for any partition makes the partition tables corrupt and requires an e2fsck to fix. This has happened to me on two computers, one has 2x 74Gb raptors split into 100Mb RAID1 for /boot, 15Gb RAID0 for / and the rest RAID0 for /home and one has 4x WD640s split into 100Mb RAID1 for /boot, 15Gb RAID0 for / and the rest RAID5 for /home. I have had to recover the /home partitions on both using a Live Desktop version because despite selecting to not format them, the following screen still says the partition tables are due to be changed.

Revision history for this message
NickJ (nickd-jones) wrote :

I've just gone through the setup routine for Karmic and on the page after confirming the computer not to boot with degraded RAID, it doesn't warn against any partitions having their structure changed and only says that / and /boot are due to be formatted. This contrasts to Lucid where it warns /, /boot and /home will have their partition tables changed, and also mentioning that / and /boot will be formatted. Also noted using the Live Desktop CD that the new / and /boot partitions created by Lucid's setup routine ARE readable.

Karmic ISO: ubuntu-9.10-alternate-amd64.iso

Lucid ISO: ubuntu-10.04-rc-alternate-amd64.iso

At the moment, the only way I have to get a clean install of Lucid onto my PC is to clean install Karmic first then upgrade.

Revision history for this message
Colin Watson (cjwatson) wrote :

You might find that this is fixed by the next server build, as this sounds awfully like bug 542210.

Revision history for this message
stop (whoopwhoop) wrote :

Hhm, I'm not entirely sure this is a duplicate (although it seems along the same lines). I'll grab the daily build and test it, to confirm...

Revision history for this message
Alvin Thompson (alvint-deactivatedaccount) wrote :

NickJ, your problem is actually bug 191119 or bug 568183 (the latter is currently a dup of the former).

Revision history for this message
stop (whoopwhoop) wrote :

Well in any case, I am not getting the errors I was getting any more...

Revision history for this message
NickJ (nickd-jones) wrote :

Alvin I have no idea why you're so adamant this is related to bug 191119, both me an whoop have said the patch in bug 542210 have fixed our problems. 191119 and 568193 don't even appear related.

Revision history for this message
stop (whoopwhoop) wrote :

Well, I can't be certain it's the patch from bug 542210 that fixed it for me :-). All I can say is it isn't broken any more...

Revision history for this message
Alvin Thompson (alvint-deactivatedaccount) wrote :

Hi, NickJ,

If I interpret Whoop correctly, his problem was that he tried to install Ubuntu to 2 separate machines while leaving his RAID /home partition alone on each. The end result was the installer goofed his install and gave him that "failed to remove conflicting files" message, but in both cases his /home RAID partition was undamaged. His end result was a botched install, on the first machine to sda1, and on the second machine to md0(?), but neither RAID /home directory had any problems.

Whoop, if this is not the case, please let me know.

In your case, NickJ, you also tried to install Ubuntu while leaving your RAID /home partition alone, but things diverge soon after that.

1. In your case, although you elected to not format or modify your RAID /home partition, the installer told you that the partition tables would be changed, anyway. Whoop makes no mention of this in his case. And he has screen shots of his install which also don't show anything abnormal.

2. In your case, the installer actually corrupted your RAID /home partition, despite you telling the installer not to touch it. This did not happen in Whoop's case.

This next part gets logically convoluted, but bear with me. You both started off with similar starting positions, and you both used similar procedures, but you had markedly different results. Therefore, there must be a different bug at play in your case. If you also received the "failed to remove conflicting files" error, it logically follows that you must have been blessed with two bugs. Lucky you!

Additionally, the results you mention in 1 and 2 above are cornerstones of the bugs I mentioned. See that first long email in bug 568183, which is currently a duplicate of bug 191119.

Long story short: it was likely *you* who fixed your second bug when you rebuilt your corrupted RAID arrays, and the fix to this bug fixed your remaining problem. BTW, how did you wind up recovering your RAID drives?

Revision history for this message
stop (whoopwhoop) wrote :

That's close enough for me Alvin :-)

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.