[clug] dist-upgrade to Debian Lenny - flagged boot problem

[Daniel Pittman]

Felix Karpfen <felix.karpfen at gmail.com> writes:
> Daniel Pittman wrote:
>> Felix Karpfen <felix.karpfen at gmail.com> writes:
>>> Daniel Pittman wrote:
>>>> Felix Karpfen <felix.karpfen at gmail.com> writes:
>>>>> The Debian Lenny Release Notes include the following entry:
>>>>>
>>>>>  4.8. System boot hangs on Waiting for root file system
>>>>>  Procedure to recover from /dev/hda that became /dev/sda
>>>> [...]
>>>>
>>>>> Given that only "some users" encounter this problem, what do I have to do to
>>>>> not fall into this "some user" subdivision?
>>
>>
>> Anyway, moving to some physically independent solution to mounting is a good
>> idea; recent kernels are moving more and more to asynchronous probing, which
>> means that *if you have multiple physical HBAs[1] you will *NOT* have a
>> predictable device detection order over boot*.
>
> I <believe|hope> that this is the answer to my original question.

Are you effected?  No, it isn't an answer to that.

> I use a desktop computer which has served me adequately for some 5 years!
> HBAs and Fibre Channel loops (after checking out what they are on the
> Internet!) are well and truly over my horizon.

Sorry, I shouldn't have used the "HBA" jargon.  That is just a "host-bus
adapter", of which a SATA or PATA controller is an example.  Not often called
one, perhaps, but still. :)

Um, the key problem in the Debian upgrade, in fact, is that some former PATA
controllers (/dev/hd?) are now run through libata, which is a SCSI driver for
them, so they change to /dev/sd?.

(also, it means that you are /not/ immune to this just because you don't have
 esoteric, Enterprise hardware.  sorry.)

>>> While I have now located the needed details for UUID mapping, the release
>>> notes - which assume that I use GRUB to boot - do not give me any guidance
>>> on what needs changing in /etc/lilo.conf.
>>
>> Well, assuming you use the stock initrd, "root=UUID=whatever" in the kernel
>> command line is all you need.
>
> If that were the whole story, I might not have written for advice.  But I was
> under the impression that the UUID mapping also needs to be extended to all
> the entries in /etc/fstab.  On my setup, these are numerous!

Er, that question was about lilo.conf, which I answered in the context of.

The answer to the other question, which is about fstab updates, is:

The 'LABEL=foo' or 'UUID=foo' syntax is standard, though; to fetch the
identifiers of any filesystem you can use:

    ls -l /dev/disk/by-uuid/* /dev/disk/by-label/*
    # ...or...
    blkid /dev/sdc1

Replace the device entries in fstab with those; from my own:

    UUID=bfcef08e-c00d-42a3-a9d6-9c915aabab54 /boot ext3 defaults,noatime ...

[...]

> There is a great deal more info in the reply that is well beyond my current
> <understanding|competence>.  But if my assumption - flagged above - is
> correct, then I believe that I can safely remain in my state of blissful
> ignorance.

Sorry, no. :(

> Again, thank you for taking the time to help me to make sense of the Debian
> release notes.

That's OK.  This is one of those painful transitions that /might/ be a big
problem, so caution is well advised.

        Daniel

-- 
✣ Daniel Pittman            ✉ daniel at rimspace.net            ☎ +61 401 155 707
               ♽ made with 100 percent post-consumer electrons
   Looking for work?  Love Perl?  In Melbourne, Australia?  We are hiring.