Custom Query (684 matches)

Hardware: Dell PowerEdge 1855 (Blade server) OS: Fedora Core 5 (LVM based) Boot options: added "ramdisk_blocksize=1024" (fedora specific)

All blades in teh chasis are exactly the same. After install on one blade, we imaged a DVD using the most recent Mondo. Using this DVD, we attempted to restore onto another of the blades. This crashed out after it started to format the LVM partitions it just created.

This was first attempted on non-interractive. We then tried on interractive, then tried to mkfs.ext3 on the LVM partitions from expert mode. Each resulted in:

Writing INODE tables: bad page state in process 'mkfs.ext2'

we could not get log files as it locked fully and needed reboot. Other errors were displayed on screen, but after about 30 seconds to 1 minute, the following was displayed:

Bug: spinlock lockup on CPU#0,mkfs.ext2/7559,c06c9380 (tainted: G B)

If we can provide further information, please ask.

Chris McLean

The process of creating the restoration disks should offer the following capabilities:

1. A pre-execution script. This will be run *before* the restoration starts. This should be started automatically, if the CD was built with "-H". If not, the user should be prompted to run fully automatic restore script. This typically will be used to custom-partition the disks, perform any complex soft raid/LVM setup needed. This should also be able to edit CD files such as the mountlist.txt. This script should be able to start the scriptable restoration script described next.

2. Mondorestore should have a scriptable invocation method. This can be accomplished using an answer file, or at least (maybe that's better) by offering command line options for each question asked during restoration. The target is to perform restoration from a script with full option control.

3. post-restore script functionality should be provided. Thinking about this, and if the pre-script is the one that starts mondorestore with options/answer-file, then the postscript might not be needed, as the pre-script will continue after mondorestore has done its magic! So the functionality is there.

Would be academic except for these circumstances -

1. Ubuntu - which points sh to /bin/dash (​https://wiki.ubuntu.com/DashAsBinSh) together with

2. http://trac.mondorescue.org/wiki/TroubleShooting - Trouble-Shooting mindi

Using sh -x /usr/sbin/mindi 2>&1 gives different results than mindi by itself.

example:

$ ./fragment
TMPMODPROBE_FLAG=Y

$ sh -x ./fragment  2>&1 
+ [ -d /etc/modprobe.d ]
+ cut -c1-3
+ uname -r
+ [ 2.6 == 2.6 ]
[: 1: ==: unexpected operator
+ echo TMPMODPROBE_FLAG=N
TMPMODPROBE_FLAG=N

(Yes, if one is observant one see the 'unexpected operator' message, but the script continues to run, so I was surprised by the result.)

$ cat fragment
#!/bin/bash 
# Test mindi fragment under debug - sh -x 
if [ -d "/etc/modprobe.d" ] && [ `uname -r | cut -c1-3` == "2.6" ] ; then
        echo TMPMODPROBE_FLAG="Y"
else
        echo TMPMODPROBE_FLAG="N"
fi

If the == is changed to = then the result is -

$ sh -x ./fragment  2>&1 
+ [ -d /etc/modprobe.d ]
+ uname -r
+ cut -c1-3
+ [ 2.6 = 2.6 ]
+ echo TMPMODPROBE_FLAG=Y
TMPMODPROBE_FLAG=Y

Of course, this could be solved also by changing to -

$ bash -x ./fragment  2>&1 
+ '[' -d /etc/modprobe.d ']'
++ uname -r
++ cut -c1-3
+ '[' 2.6 == 2.6 ']'
+ echo TMPMODPROBE_FLAG=Y
TMPMODPROBE_FLAG=Y

So maybe the troubleshooting tip should have bash -x /usr/sbin/mindi 2>&1 ?

Regards,

Scott