Discussion:
FAI on a thumper
Carsten Aulbert
2010-02-26 07:58:37 UTC
Permalink
Hi all

we are currently trying FAI 3.3.3 on a thumper (Sun Fire x4500) but right now
it does not like it too much:

[...]
Executing: parted -s /dev/sdy unit chs print free
Executing: parted -s /dev/sdz unit TiB print
Executing: parted -s /dev/sdz unit B print free
Executing: parted -s /dev/sdz unit chs print free
Executing: vgdisplay --units m -s
Executing: mdadm --detail --scan --verbose -c partitions
INTERNAL ERROR in setup-storage:
Invalid device /dev/sdy1
Please report this error to the Debian Bug Tracking System.

the device /dev/sdy is there (I can query it via smartctl) but sdy1 is not
(yet) there. If I run against another disk, e.g. sdc all is fine, but this
machine can only boot from very specific disks :(

When enabling debugging the output get *very* lengthy, but I can try to
capture it, if people would like to see it.

Any ideas how I can proceed?

Cheers

Carsten
Alexander Bugl
2010-02-26 08:19:35 UTC
Permalink
Hi Carsten!
Post by Carsten Aulbert
we are currently trying FAI 3.3.3 on a thumper (Sun Fire x4500) but right now
Executing: parted -s /dev/sdy unit chs print free
Executing: parted -s /dev/sdz unit TiB print
Are you sure about sdy?

According to http://docs.sun.com/source/820-4893-13/rhel.html Table 4-1
it depends on the firmware version and on Virtual Floppy / Virtual
CD-ROM being mounted, so you should probably try sdx ...

(And you should change sdz to sdab or sdac ...)

I have no experience with Thumpers and FAI, but spent some time using
kickstart for CentOS to find the correct boot devices -- and in my case
it has been sdy and sdac, even with Firmware 2.0.2.5 and no Virtual
Floppy / Virtual CD-ROM, so in my case the table was wrong. Probably it
depends on the kernel version used ...

With regards, Alex
--
Alexander Bugl, Central IT Services, ZMAW
Max Planck Institute for Meteorology
Bundesstrasse 53, D-20146 Hamburg, Germany
tel +49-40-41173-351, fax -356, room PE048
Carsten Aulbert
2010-02-26 09:22:44 UTC
Permalink
Resending without the attachment (Thomas, 15kB is REALLY strict)

Hi Alex,
Post by Alexander Bugl
Are you sure about sdy?
I thought so, as I looked before where no GPT label was
Post by Alexander Bugl
According to http://docs.sun.com/source/820-4893-13/rhel.html Table 4-1
it depends on the firmware version and on Virtual Floppy / Virtual
CD-ROM being mounted, so you should probably try sdx ...
(And you should change sdz to sdab or sdac ...)
That was just copy paste from the FAI output and FAI searches /dev in a
special order, i.e.

sda, sdaa, sdab, sdac, ..., sdb, sdc, sdd, ...
Post by Alexander Bugl
I have no experience with Thumpers and FAI, but spent some time using
kickstart for CentOS to find the correct boot devices -- and in my case
it has been sdy and sdac, even with Firmware 2.0.2.5 and no Virtual
Floppy / Virtual CD-ROM, so in my case the table was wrong. Probably it
depends on the kernel version used ...
*sigh* ;)

But the hint did not help:

INTERNAL ERROR in setup-storage:
Invalid device /dev/sdx1
Please report this error to the Debian Bug Tracking System.

Interestingly:
root at s01:/# ls -l /dev/sdx*
brw-rw---- 1 root disk 65, 112 Feb 26 09:04 /dev/sdx
brw-rw---- 1 root disk 65, 113 Feb 26 09:04 /dev/sdx1
brw-rw---- 1 root disk 65, 121 Feb 26 09:04 /dev/sdx9

and more so:
root at s01:/# parted /dev/sdx print
Model: ATA HITACHI HDS7250S (scsi)
Disk /dev/sdx: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt

Number Start End Size File system Name Flags
1 131kB 500GB 500GB zfs

root at s01:/# parted /dev/sdy print
Model: ATA HITACHI HDS7250S (scsi)
Disk /dev/sdy: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number Start End Size Type File system Flags
1 32.3kB 263MB 263MB primary ext3 boot
2 263MB 26.5GB 26.2GB primary ext3
3 26.5GB 60.8GB 34.4GB primary linux-swap
4 60.8GB 500GB 439GB extended lba
5 60.8GB 87.0GB 26.2GB logical xfs
6 87.0GB 113GB 26.2GB logical xfs
7 113GB 115GB 2147MB logical xfs
8 115GB 500GB 385GB logical xfs

Something seems amiss here, any setup-storage gurus around (this is v1.0.3)?

Attaching the slightly edited fai.log (compressed)
EDIT: Now available under
http://atlas1.atlas.aei.uni-hannover.de/~carsten/fai-thumper.log.gz

Cheers

Carsten
Carsten Aulbert
2010-02-26 14:03:41 UTC
Permalink
OK, I think the error I hit was already fixed in SVN head:

http://svn.debian.org/wsvn/fai/trunk/lib/setup-
storage/Init.pm?op=diff&rev=5691&sc=0

Can anyone tell me, why sdt was the "old" limit?

Cheers

Carsten
Michael Tautschnig
2010-02-26 19:48:32 UTC
Permalink
Post by Carsten Aulbert
http://svn.debian.org/wsvn/fai/trunk/lib/setup-
storage/Init.pm?op=diff&rev=5691&sc=0
Well, but that fix should be in FAI 3.3.3 (which you claimed you are using)!?
Post by Carsten Aulbert
Can anyone tell me, why sdt was the "old" limit?
I think I got that somewhere from within the kernel source code, but that either
was wrong or terribly outdated.

Best,
Michael

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
Url : http://lists.uni-koeln.de/pipermail/linux-fai/attachments/20100226/354275d9/attachment.bin
Carsten Aulbert
2010-02-26 20:12:47 UTC
Permalink
Hi
Post by Michael Tautschnig
Well, but that fix should be in FAI 3.3.3 (which you claimed you are using)!?
That's what dpkg -l keeps telling me (and what's pulled in form cologne).

The nfsroot was freshly created yesterday and the web interface tells me that this was indeed fixed in 3.3.3 -
weird.
Post by Michael Tautschnig
I think I got that somewhere from within the kernel source code, but that
either was wrong or terribly outdated.
So, it was not a "qwertz" typo? ;)

Well at least now I'm stuck with grub where I have to "manually" tinker with the device map as grub seems to
like only up to 16 disks per system...

Cheers

CArsten
Michael Tautschnig
2010-02-26 20:24:24 UTC
Permalink
Post by Carsten Aulbert
Hi
Post by Michael Tautschnig
Well, but that fix should be in FAI 3.3.3 (which you claimed you are using)!?
That's what dpkg -l keeps telling me (and what's pulled in form cologne).
The nfsroot was freshly created yesterday and the web interface tells me that this was indeed fixed in 3.3.3 -
weird.
I just checked the contents of the packages from cologne, indeed the fix is in
there. Do you also have the koeln repository in /etc/fai/apt/sources.list? AFAIK
this is what make-fai-nfsroot uses for installing packages into the NFSROOT.
Post by Carsten Aulbert
Post by Michael Tautschnig
I think I got that somewhere from within the kernel source code, but that
either was wrong or terribly outdated.
So, it was not a "qwertz" typo? ;)
Nope.
Post by Carsten Aulbert
Well at least now I'm stuck with grub where I have to "manually" tinker with the device map as grub seems to
like only up to 16 disks per system...
Are you using grub-pc? I'd hope that at least grub 2 supports more than 16
disks... Then, of course, our example GRUB_PC class is still very
experimental...

Best,
Michael

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
Url : http://lists.uni-koeln.de/pipermail/linux-fai/attachments/20100226/33d8904b/attachment.bin
Carsten Aulbert
2010-02-27 10:18:30 UTC
Permalink
Hi Michael,
Post by Michael Tautschnig
I just checked the contents of the packages from cologne, indeed the fix is
in there. Do you also have the koeln repository in
/etc/fai/apt/sources.list? AFAIK this is what make-fai-nfsroot uses for
installing packages into the NFSROOT.
Well, using FAI for a couple of years now and then this PEBKAC - sorry yes,
the nfsroot was built with the old version from Debian.

I've added the K?ln repo and now it's fixed automatically in the NFS root.

Thanks for the hint!

Cheers

Carsten

PS: If I find some time I might experiment with "grub2"/pc-
boot/whatevertheywillcallitnext ;)

Carsten Aulbert
2010-02-26 20:35:03 UTC
Permalink
Hi
... not entirely on topic, besides that we would love to be able
we tried running linux on a thumper some time ago and it sucked
badly. the raid drivers were just not ready for production and
performance was bad on top.
I do not know yet, but so far everything seem to be as bad (IMHO). Solaris
10u8 crashes hard when a mere 20 clients try to read or write large amounts of
data from the system, if I go to a 10 Gb/s NIC the network stacks dies....

Opensolaris was not better, Nexenta crashed also, kfreeBSD does not have zfs
tools, well now it was our time for Debian/Thumper...

Cheers

Carsten
Continue reading on narkive:
Loading...