FW: (Scyld) ext2 fs error when writing beoboot image to slave HDD

[Eric T. Miller]

-----Original Message-----
From: beowulf-admin at beowulf.org [mailto:beowulf-admin at beowulf.org]On
Behalf Of Sean Dilda
Sent: Monday, August 06, 2001 4:02 PM
To: Eric T. Miller
Cc: beowulf at beowulf.org
Subject: Re: (Scyld) ext2 fs error when writing beoboot image to slave
HDD


On Sun, 05 Aug 2001, Eric T. Miller wrote:

>> Man!  I am so close!  Changing from the 3Com to the Dlink NIC solved all
the
>> driver issues.  Now, the slaves complete phase one and phase two boot,
>> unpartitioned drive on slaves causes error status, just like
documentation
>> said it would.  Creating default partition tables and writing those to
the
>> slaves goes off with out a hitch.  But, when I envoke :
>>
>> 		beoboot-install -a /dev/hda
>>
>> to write the boot image to the slave hard drive, the script executes
>> normally, but ends with an error:
>>
>> 		ext2fs check_if_mount no such file or directory while determining
whether
>> /dev/hda1 is mounted
>>
>> followed by some other errors that relate to the same.
>
>Is the hard drive in your slave node an IDE drive that is the master on
>the primary IDE chain?  If not, then 'hda' is not correct for your drive
>and would explain why you are getting this error.>>>


Yes, HDD in all five slave nodes are primary/master
>
>Also, did you reboot the slave nodes after you modified the partition
>tables?  Sometimes you have to reboot the machine so that the kernel can
>read the new partition table.

No, I had not done that.  I went through the procedure again, this time
rebooting after writing the partitions to the slaves.  After the reboot, I
continue with the install of the boot image (beoboot-install -a /dev/hda)
and I get the same results as noted above.  Is it possible that my FSTAB is
wrong?  I used the defaults .  Also, when I check the partitions written to
the slave nodes, /dev/hda1 is only 8MB?  Is that right?

The nodes are all appearing in Beo Status, as up, but unavailable.  The
processor usage, memory and network usage are reading good values, the disk
value is 100% of zero.

Would trying diskless be a good solution for me?  Is that difficult to
configure?  This cluster is for demonstration purposes, so I am just
interested in getting it running the simplest aproach available.
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ATT00016.dat
Type: application/pgp-signature
Size: 232 bytes
Desc: not available
Url : http://www.scyld.com/pipermail/beowulf/attachments/20010806/737bd4f5/ATT00016.bin