[linux-lvm] Re: LVM2 on RAID

Luca Berra bluca at comedia.it
Wed Apr 27 16:27:27 UTC 2005


On Wed, Apr 27, 2005 at 03:03:47PM +0200, Måns Rullgård wrote:
>> /* start quote */
>>
>> From: Luca Berra <bluca comedia it> 
....
>> anyway the major issues are:
>> 1) raid autodetect uses only the minor number stored in the superblock
>> to identify the array components, move a disk from a different machine
>> to yours and reboot it to enjoy the show
>
>Is this really true?  My kernel log has messages like these:
>
>md: Autodetecting RAID arrays.
>md: autorun ...
>md: considering sdb3 ...
>md:  adding sdb3 ...
>md: sdb1 has different UUID to sdb3
>md:  adding sda3 ...
>md: sda1 has different UUID to sdb3
>md: created md2
>
>If it uses the UUID, there should be little risk of misdetection.

the issue is: if it finds the wrong one before the right one

>A problem I have run into with LVM, was disks from different machines,
>using the same VG names.
>
very true.

>> Can anyone conclude that passing the parameter md=0,/dev/* to kernel
>> parameter is a better approach than using auto detect? I don't have
>> a problem with having a rather long parameters in my
>> grub.conf/lilo.conf.
>
>As usual, both ways have their own advantages and disadvantages.  What
>I like about autodetection, is that things just work, even if I
>shuffle my disks for some reason.  
>
what i did is write mdassemble to be used in initrds
it uses a mdadm.conf file where you can put the UUID and have all disk
that match a specific UUID be assembled in your MD

L.

-- 
Luca Berra -- bluca at comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \




More information about the linux-lvm mailing list