old-style fstab not updated during system upgrade

[Nils Pickert]

When upgrading using update-manager -c -d from dapper to feisty my fstab which had still /dev/hda1-sytle entries in it was not updated to the /dev/disk/by-uuid system... As feisty seems to rename my disk from hda to sda this lead to a failure during boot, and I couldn' mount the home partition...

[Jim Gettys (jg-laptop)]

I too suffered from this; complete obscure failure; thankfully, it was my home partition and not by root, so after an hour or so, I figured it out and am running.

This *REALLY* needs to get fixed.
                                - Jim Gettys

[Jim Gettys (jg-laptop)]

<cjwatson> don't upgrade direct from dapper to feisty though
<jg> upgrade was from Edgy.
<cjwatson> perhaps you could help us debug that with reference to /var/lib/dpkg/info/volumeid.postinst, which is supposed to do that migration?
<jg> cjwatson: there is no case for /dev/hda? in the case statement....
<cjwatson> there doesn't need to be, it's included in /dev/*
<StevenK> if dpkg --compare-versions "$2" lt "093-0ubuntu5"; then
<jg> cjwatson: my shell is too rusty to help much... I've grown pointy hair...
<StevenK> That seems to strike me as backwards, and I don't know why.
<robertj> is the release officially out now? tis on the top of /.
<cjwatson> "if version we're upgrading from is older than 093-0ubuntu5, then do the migration"
<jg> cjwatson: here's the scenario, I think that caused me to lose....
<pitti> hm, should really be lt-nl, but that shouldn't cause this bug
<jg> I think checking the version is a *bad* idea.
<pitti> or, no, maybe not
<jg> I had a flaky motherboard, but thought it was my disk.
<jg> I got file system damage.
<jg> I went and got another disk.
<cjwatson> the problem is that if you don't do that then the migration happens over and over and over again
<cjwatson> and people who have some problem and really need to go back to mount-by-something-else lose
<jg> Did a fresh install of Edgy, and then moved my /home back.
<jg> using the "normal" old fashioned syntax.
<cjwatson> could you drop this in the bug? The udev maintainer is on holiday today, so information on IRC is liable to get lost
<jg> I think you need to see if /dev/hda? exists, and go from there, rather than key it on version.
<jg> ok, I'll add to the bug.

[Kyle Schneider (kschneider2003)]

I ran into this as well on an upgrade from edgy. I had old style fstab entries and had to deal with identifying which sd* devices those entries corresponded to. Perhaps the fstab should be parsed looking for uncommented /dev/hd* entries and convert only those entries?

[alexey (alexey-wmd)]

I had the same problems during my upgrade to feisty. Device names was renamed (hda -> cdrom, hdc -> sda, sda -> sdb) so I had to fix fstab to get my system work.
Here's my old fstab:

proc /proc proc defaults 0 0
/dev/sda1 / ext3 defaults,errors=remount-ro 0 1
/dev/sda3 /home ext3 defaults 0 2
/dev/sda4 /home/media xfs defaults 0 2
/dev/hdc1 /media/hdc1 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/hdc5 /media/hdc5 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/hdc6 /media/hdc6 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/hdc7 /media/hdc7 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/hdc8 /media/hdc8 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/hdc9 /media/hdc9 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sda2 none swap sw 0 0
/dev/hda /media/cdrom0 udf,iso9660 user,noauto 0 0

fixed fstab:

proc /proc proc defaults 0 0
/dev/sdb1 / ext3 defaults,errors=remount-ro 0 1
/dev/sdb3 /home ext3 defaults 0 2
/dev/sdb4 /home/media xfs defaults 0 2
/dev/sda1 /media/hdc1 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sda5 /media/hdc5 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sda6 /media/hdc6 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sda7 /media/hdc7 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sda8 /media/hdc8 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sda9 /media/hdc9 vfat defaults,utf8,umask=007,gid=46 0 0
/dev/sdb2 none swap sw 0 0
/dev/cdrom /media/cdrom0 udf,iso9660 user,noauto 0 0