Re: Disk /dev/md6 doesn't contain valid partition table
Hi!
On Thu 06 November 2008 02:37, Francesco Pietra wrote:
> Hi all:
> The story continues, unfortunately.
>
> After the last kind mail below, I could carry out several long
> parallel computations on all 4 physical (8 logical) dual-opteron 875
> processors , or a fraction of them(amd6a lenny; H8QCE Supermicro
> motherboard; two WD Raptor HD (ext3 file syst) under Linux raid1 as
> previously described; 24GB ram registered ECC Kingston 400 MHz). The
> last computation with 8 processor finished regularly and the computer
> was also regularly shut down (shutdown -h now). No overheating during
> the computations, the machine is very well ventilated and there is
> partial control from "sensors" (Supermicro never provided data to
> fully set up sensors). The booting procedure is set up so as to end at
> the terminal screen. If I want the X sytem (which I use on rare
> occasions) I have to command as user "startx".
> =================================
>
> This morning, trying to start the computer, the kernel seems to be
> loaded but the procedure did not end successfully. On the last part of
> the screen output:
>
> RPB 000...
> R10 000..
> IP13 000...
> FS
> CS
> CR2
> DR0 000 DR1 000 DR! 000
> DR3
>
> Call Trace
> ffff.. down_read_trylock
> ffff... do_page_default
> fff... vma_link
> fff... error_exit
> fff clear_user
> fff pad_zero
> fff load_elf_library
> fff get_arg_page
> fff .. copy_strings
> fff ... search_binary_handler
> fff.. do_execve
> fff.. stub_execve
>
>
> As the procedure did no proceed farther than this, the computer was
> reset from the "RESET" button, whereby re-booting occurred and the
> screen - unlike the above - showed (inter alia):
>
> PCI found disabled HT MSI Mapping on 0000...
>
> Assign interrupt mode Found MSI capability
>
> Real Time Clock Driver v1.12ac
>
> input MacIntosch mouse button emulation
>
> Freeing unused kernel memory: 316k
>
> input: AT translated Set 2 keyboard as /class/input/input1
>
> (the procedure halted here. Retrying, same sequence of events.
> ===========================
>
>
> Knoppix 5.3 loaded correctly, as far as I could see. All memory and
> all 8 logical processors and the MacIntosh type mouse were loaded.
> However, clicking on either "sda" or "sdb" on the screen:
>
> Could not mount. Could not determine the file system type and no one
> was specified.
Francesco,
You would not be able to mount either /dev/sda or /dev/sdb. They designate
your entire hard drives, not valid partitions. Looking back to your earlier
posts, your output from `cat /proc/mdstat` shows that your valid partitions
are: sd[ab]1, sd[ab]2, ..., sd[ab]8. You should be able to mount any one of
these partitions.
If you do mount one of these partitions, you should it as read-only, otherwise
you'll upset your raid-1 device. If you want write capability, then I would
use mdadm to start the appropriate raid device (md0, md1, ..., md7) and then
mount the raid device.
> On the terminal, /home is knoppix. Cd to /mnt showed sda and sdb but
> on cd to these two directories, ls did not show anything. (Problem
> with the HDD?). The computer was shut down from Knoppix.
Knoppix is just confused about your partitions. I can't remember, but are the
partition types set to 'Linux Raid Autodetect'?
> Waiting for some general advice or specific direction as to what
> should be checked.
When trying to diagnose Linux software raid, mdadm is your other friend. Think
of it as a Swiss Army knife. With it, you can create, assemble & manage md
devices. It's pretty intimidating at first, but not that difficult to use.
Check out: http://tldp.org/HOWTO/Software-RAID-HOWTO.html
> Thanks
>
> francesco
>
>
> I
HTH
cmr
> On Thu, Oct 30, 2008 at 6:58 PM, C M Reinehr <cmr@amsent.com> wrote:
> > Francesco,
> >
> > On Thu 30 October 2008 04:14, Francesco Pietra wrote:
> >> Then, let me ask about a raid1 devised mainly for the system than the
> >> data (once a calculation is finished, all files are sent to a desktop;
> >> no data remain in my home. This also because I need fast HDDs, while
> >> space on WD Raptor is very limited: 150GB each HDD). With so many
> >> applications that require a long compilation, a fault in the system is
> >> a terrible waste of time. It occurred to me, at the time I had raid1
> >> with two cheap HDDs. that one HHD suffered mechanical failure.
> >> Replaced, the system on raid1 was automatically restored.
> >>
> >> Which one of the kindly offered recipes for raid1 is the most suited
> >> to the above case? (that, I guess, is quite common - if not the norm -
> >> in computational chemistry, biology, etc).
> >
> > From your point of view, I don't think it matters. Each method, with or
> > without LVM, is equally reliable. The advantage of LVM is just that it
> > gives you more administrative flexibility. Should you subsequently decide
> > that one volume (partition) is too small and another too large, you can
> > resize them on the fly--rather than having to backup the partitions,
> > resize them, & then restore the contents. Also, as someone on another
> > group observed. Raid-1 provided redundancy and LVM adds striping.
> >
> > Cheers!
> >
> > cmr/anonymous
> >
> >> Thanks
> >> francesco
> >>
> >> On Thu, Oct 30, 2008 at 4:01 AM, Alex Samad <alex@samad.com.au> wrote:
> >> > On Wed, Oct 29, 2008 at 11:07:36AM -0500, C M Reinehr wrote:
> >> >> Francesco,
> >> >>
> >> >> On Wed 29 October 2008 06:16, Francesco Pietra wrote:
> >> >> > On Wed, Oct 29, 2008 at 11:40 AM, Alex Samad <alex@samad.com.au>
wrote:
> >> >> > > On Wed, Oct 29, 2008 at 08:24:55AM +0100, Francesco Pietra wrote:
> >> >> > >> On Wed, Oct 29, 2008 at 7:06 AM, Douglas A. Tutty
> >> >> > >> <dtutty@vianet.ca>
> >> >>
> >> >> wrote:
> >> >> > >> > On Wed, Oct 29, 2008 at 05:44:31AM +0100, Francesco Pietra
> >> >> > >> > wrote:
> >> >
> >> > [snip]
> >> >
> >> >> PS I agree with Alex regarding LVM2. I have only two partitions
> >> >> defined on my hard drives, one each for two md arrays. The first md
> >> >> device is for my boot partition. The second for everything else. The
> >> >> everything else, then, is managed by LVM2 with logical volumes for
> >> >> each seperate file system. LVM2 is a little intimidating but once up
> >> >> & running is much easier to manage.
> >> >
> >> > I usually go with 3
> >> >
> >> > 1 - 500M /boot
> >> > 2 - 20G for /
> >> > 3 - the rest for lvm.
> >> >
> >> > I like keeping the / fs on something simple especially if I have to
> >> > rescue it
> >> >
> >> >> --
> >> >> Debian 'Etch' - Registered Linux User #241964
> >> >> --------
> >> >> "More laws, less justice." -- Marcus Tullius Ciceroca, 42 BC
> >> >>
> >> >>
> >> >> --
> >> >> To UNSUBSCRIBE, email to debian-amd64-REQUEST@lists.debian.org
> >> >> with a subject of "unsubscribe". Trouble? Contact
> >> >> listmaster@lists.debian.org
> >> >
> >> > --
> >> > "I need to be able to move the right people to the right place at the
> >> > right time to protect you, and I'm not going to accept a lousy bill
> >> > out of the United Nations Senate."
> >> >
> >> > - George W. Bush
> >> > 10/31/2002
> >> > South Bend, IN
> >> >
> >> > -----BEGIN PGP SIGNATURE-----
> >> > Version: GnuPG v1.4.9 (GNU/Linux)
> >> >
> >> > iEYEARECAAYFAkkJI6IACgkQkZz88chpJ2MI7gCg0lbsxErdXiAigCJX5IExjEQe
> >> > A4QAnjCjRrIskHRn/YW8nynxFnXBBa2Y
> >> > =V9OO
> >> > -----END PGP SIGNATURE-----
> >
> > --
> > Debian 'Etch' - Registered Linux User #241964
> > --------
> > "More laws, less justice." -- Marcus Tullius Ciceroca, 42 BC
> >
> >
> > --
> > To UNSUBSCRIBE, email to debian-amd64-REQUEST@lists.debian.org
> > with a subject of "unsubscribe". Trouble? Contact
> > listmaster@lists.debian.org
--
Debian 'Etch' - Registered Linux User #241964
--------
"More laws, less justice." -- Marcus Tullius Ciceroca, 42 BC
Reply to: