Bugs item #1068036, was opened at 2004-11-17 14:19
Message generated for change (Comment added) made by laudares
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=109368&aid=1068036&group_id=9368
Category: Installation
Group: 4.0
Status: Open
Resolution: None
Priority: 9
Submitted By: Thomas Naughton (naughtont)
Assigned to: Bernard Li (bernardli)
Summary: Fedora (FC2) - scsi HD bug
Initial Comment:
When testing with oscar-4.0 material on Fedora Core 2
(x86) with a SCSI HD I had problems after the node
reboots.
The node seems to build properly (step6) but on the
next reboot after the SI install you get a Kernel PANIC
can't find root (sda6). The sda6 partition exists and
appears to have the correct data, ie. it is mountable
by hand from a rescue disk.
I/we tried playing with a few Grub options to no avail.
Bernard had similar issues (kernel panic can't mount
root) on some of his systems and it seemed to be due to
missing entries (apparently un-needed) in modules.conf
(2.4 kernel) and therefore were missing in the new
modprobe.conf (2.6 kernel). I do not have the test
system available to confirm this portion for our issue.
Note, the headnode had FC2/SCSI but was built by hand
from CD's and worked just fine.
REPRODUCE PROBLEM:
+ use oscar-4.0beta tarball on Fedora Core 2 system
with SCSI harddrives.
+ After building nodes, upon reboot, node(s) should
have kernel panic
----------------------------------------------------------------------
Comment By: Fernando Laudares Camargos (laudares)
Date: 2004-12-02 21:56
Message:
Logged In: YES
user_id=931808
Yes, a SCSI Maxtor ATLAS10k4
(2.4.21-15.EL)
----------------------------------------------------------------------
Comment By: Nobody/Anonymous (nobody)
Date: 2004-12-02 21:32
Message:
Logged In: NO
We have tried this on a head node that is identicall to our
client node, and still have the problem.
I think it is becuase systemimager or more specfically 2.4.25
uses /dev/hdX (we have serial-ata disks) and 2.6
uses /dev/sdX for the serial drives. When you say you have
scsi drive do you have real scsi or serial ata?
I updated my initrd image from a working system, and then
the kernel was able to see root, but I recieve a panic when
the kernel tries to mount / to load init. I have updated
grub.conf, and /etc/fstab as well as the initrd. I have no idea
why the system won't boot after I have updated the files.
The kernel complians about an ext3 error when trying to
mount root, I upgraded my partitions from ext2 to ext3 and I
still get the same error which now results in the kernel not
finding init. One more step to get through. Oh yes, I also
update modprobe.conf
Jason
[EMAIL PROTECTED]
----------------------------------------------------------------------
Comment By: Fernando Laudares Camargos (laudares)
Date: 2004-12-02 21:30
Message:
Logged In: YES
user_id=931808
We had this same problem with RHEL 3 (ia64) and that was due
to the fact that the system was recognizing our cd-rom as
the hd... To correct this we had to modify the DISKORDER
from the oscarimage.master script
(/var/lib/systemimager/scripts)
from: echo DISKORDER=${DISKORDER=hd,sd,cciss,ida,rd}
to: echo DISKORDER=${DISKORDER=sd,hd,cciss,ida,rd}
----------------------------------------------------------------------
Comment By: Nobody/Anonymous (nobody)
Date: 2004-12-02 21:22
Message:
Logged In: NO
Our head node is the same as our client nodes, and we have
this problem. We are using serial ata actually. I think that
the systemimager (2.4.25) uses the device as /dev/hdX and
the newer kernels use the device as /dev/sdX....
After updating the initrd on our client image, with one that
supports scsi, the kernel is able to find root, but it then can't
mount /root to find init. I updated the /etc/fstab on the
client but this did not seem to help. Can I build an image
from a working system with oscar? ( I know how to do this
with system imager, but oscar does not give me the option to
select that image when defining clients.)
Jason
[EMAIL PROTECTED]
----------------------------------------------------------------------
Comment By: Nobody/Anonymous (nobody)
Date: 2004-12-02 13:08
Message:
Logged In: NO
I think this is related to the same errors seen under the pvfs
bug. I had to copy over a working initrd to my systemimager
image directory in order to get the image to boot. It looks
like something is haveing a problem building the initrd
properly. The initrd that was built by default did not have the
scsi modules in it.
Jason
[EMAIL PROTECTED]
----------------------------------------------------------------------
Comment By: Bernard Li (bernardli)
Date: 2004-11-30 19:13
Message:
Logged In: YES
user_id=879102
- code checked into SVN
- release notes will be updated
- SVN r2764
----------------------------------------------------------------------
Comment By: Bernard Li (bernardli)
Date: 2004-11-26 22:15
Message:
Logged In: YES
user_id=879102
- I think I have working solution which is to run
systemconfigurator after /etc/modprobe.conf is generated
- I will perform some more testing before commiting it
- The user will still be able to copy over a working
modprobe.conf if they choose to
----------------------------------------------------------------------
Comment By: Bernard Li (bernardli)
Date: 2004-11-23 23:18
Message:
Logged In: YES
user_id=879102
- the fix is, if the headnode has SCSI HD, simply
copy /etc/modprobe.conf
to /var/lib/systemimager/images/oscarimage/etc (on the
headnode) and delete the
script /var/lib/systemimager/scripts/post-
install/22all.generate_modprobe_script
- this worked on a cluster that I tested on during SC04
- I will look into an alternative solution which is to hardcode
the necessary SCSI module(s) to /etc/modprobe.conf after
it's generated from /etc/modules.conf - this may work as I
have noticed that the necessary module is 'generic'
- this setup should work for both SCSI and IDE drives
- ultimately I believe this has to be fixed for SIS...
----------------------------------------------------------------------
Comment By: Nobody/Anonymous (nobody)
Date: 2004-11-22 20:37
Message:
Logged In: NO
Bernard please explain your fix, then we'll determine if
this is a 4.0 item (ie. before moving to level=9).
----------------------------------------------------------------------
Comment By: Ed Hill (edhill)
Date: 2004-11-17 16:08
Message:
Logged In: YES
user_id=79627
We also had this problem and it was because the SCSI drivers
we needed were missing from the initrd. And I *think* this
happened because our compute nodes are SCSI while the head
node is IDE.
The ways to fix it are to either build a custom kernel that
has the needed SCSI drivers compiled in or to make sure that
OSCAR has an initrd that has the necessary SCSI drivers.
So how can we specify a spicific initrd for a node or a set
of nodes (say, if some nodes are SCSI and some aren't)?
Ed <[EMAIL PROTECTED]>
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=109368&aid=1068036&group_id=9368
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://productguide.itmanagersjournal.com/
_______________________________________________
Oscar-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-devel
