Software RAID + LVM2

Fredrik Eriksson fredrik.eriksson at axis.com
Fri Jan 21 17:07:45 CET 2011


Hi, Michael

Thanks for your reply. I am trying out a few things you mentioned and 
will send you a few log files privately.


Michael Tautschnig wrote:
> Just a note: it shouldn't be necessary to add the preserve_lazy:1,2 and
> preserve_lazy:0,1 lines - setup-storage should definitely be able to figure this
> out itself, just from the preserve_lazy:vg0-home. In fact I'd even be interested
> in a try that shows that this effectively works.
>   
Ok, I've removed this and tried to install again with rebuilt nfsroot to 
remove all local changes under that directory. Again I get the blinking 
cursor in the lower left corner. I think something is done to the 
beginning of the disk since it will actually boot when loading a remote 
kernel via the PXE boot menu, but not without errors.

>> +system "/sbin/mdadm-startall";
>> # see whether there are any existing LVMs
>> &FAI::get_current_lvm;
>> +# and stop mdadm
>> +system "/etc/init.d/mdadm-raid stop";
>>
>>     
>
> Hmm, I wonder whether the kernel module md-mod is loaded before setup-storage
> starts. Could you please try to figure that out? I had hoped that if it is
> loaded, the array would be detected without such explicit starting of mdadm.
>   
I followed your lead and, instead of earlier changes, just added a lsmod 
line before get_current_lvm like this

 system "/sbin/lsmod";
 &FAI::get_current_lvm;


So I got a printout of all the modules

 md_mod                 73824  0

included, followed by

     Finding all volume groups
   No volume groups found
 (CMD) mdadm --examine --scan --verbose -c partitions 1> /tmp/9zLFVW5FMG 
2> /tmp/VtouZRXPYf
 Executing: mdadm --examine --scan --verbose -c partitions

etc and later still

 Current LVM layout
 $VAR1 = {};

which is not empty if we run mdadm-startall before get_current_lvm. 
(fai.log.lsmod in private mail)
>> # see whether there are any existing RAID devices
>> &FAI::get_current_raid;
>> @@ -177,7 +181,11 @@
>> $FAI::debug and print Dumper \%FAI::configs;
>>
>> # generate the command script
>> -&FAI::build_disk_commands;
>> +# build_disk_commands won't leave our partitions alone
>> +#&FAI::build_disk_commands;
>> +&FAI::push_command( "true", "" , "pt_complete_/dev/sda");
>> +&FAI::push_command( "true", "" , "pt_complete_/dev/sdb");
>> +
>> &FAI::build_raid_commands;
>> &FAI::build_lvm_commands;
>> &FAI::build_cryptsetup_commands;
>>     
>
> In what sense is it not leaving your partitions alone, or, rather, what does it
> break? Could you please give it another try with only the above mdadm-changes?
>   
With the build_disk_commands and without our two push_command lines, or 
vise versa, it does find the correct Current LVM layout, but exit with

 Error in task partition. Traceback: task_error task_partition task 
task_install task task_action task main
 $LOGUSER is undefined. Not saving log files to remote.
 FATAL ERROR. Installation stopped.

(fai.log.stopped in private mail)
All of the above with this disk_config

 disk_config disk1 bootable:1
 primary  -              512     -       -
 primary  -              0-      -       -
 
 disk_config disk2 bootable:1
 primary  -              512     -       -
 primary  -              0-      -       -
 
 disk_config raid
 raid1        /boot   sda1,sdb1  ext4    rw,errors=remount-ro
 raid1        -       sda2,sdb2  -       -
 
 disk_config lvm preserve_lazy:vg0-home
 vg              vg0                     md1
 vg0-swap                swap                    4096        swap        rw
 vg0-local               /usr/local              10240       ext4        rw
 vg0-src                 /usr/src                10240       ext4        rw
 vg0-usr                 /usr                    10240       ext4        rw
 vg0-var                 /var                    10240       ext4        rw
 vg0-tmp                 /tmp                    2048        ext4        rw
 vg0-home                /home                   97280       ext4        rw
 vg0-root                /                       4096        ext4        
rw,errors=remount-ro

If I use this disk_config line instead

 disk_config raid preserve_lazy:0,1 always_format:0

together with the two push_command lines and without build_disk_commands 
the installation will go through and vg0-home is preserved. 
(fai.log.vg0-home in private mail)

> What is it that gives these messages, is it grub? I think it would help me a lot
> if you could send along the full fai.log; you might want to send that in private
> mail, though.
>   
This is after grub since it does load the kernel and the initial ramdisk 
prior to that message, but I'll write a more detailed output of what is 
printed after grub

 Loading Linux 2.6.32-5-amd64 ...
 Loading initial ramdisk ...
 [   0.973192] pci 0000:01:00.0: BAR 6: no parent found for of device 
[0xfff80000-0xffffffff]
 [   0.973302] pci 0000:03:00.0: BAR 6: no parent found for of device 
[0xfffc0000-0xffffffff]
 Loading, please wait...
   Volume group "vg0" not found
   Skipping volume group vg0
 Unable to find LVM volume vg0/root
 Gave up waiting for root device. Common problems:
  - Boot args (cat /proc/cmdline)
    - Check rootdelay= (did the system wait long enough?)
    - Check root= (did the system wait for the right device?)
  - Missing modules (cat /proc/modules; ls /dev)
 ALERT!  /dev/mapper/vg0-root does not exist.  Dropping to a shell!
 (initramfs)

To clearify; This is with local changes to setup-storage, mentioned here 
and in earlier mail, but with grub preserved from previous install. This 
is probably due to the preserve_lazy at the disk_config raid line, if it 
is broken from previous installation it stays broken (the blinking lower 
left cursor).
If we use our rescue mode to load a kernel and an initrd image remote we 
get some problems with md0, but the system is quite bootable and 
vg0-home is preserved. (I'll supply /var/log/dmesg in private mail, 
there might be something useful in it, also a fai.log.437, from an 
install with FAI 4.3.7, for reference.)


Thank you, I'll be back Monday
--
Fredrik
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fredrike.vcf
Type: text/x-vcard
Size: 128 bytes
Desc: not available
URL: <http://lists.uni-koeln.de/pipermail/linux-fai/attachments/20110121/05059952/attachment.vcf>


More information about the linux-fai mailing list