FC2 - Booting from remaining raid disk

Matthew Claridge mclaridge at rwa-net.co.uk
Tue Oct 12 16:45:35 UTC 2004


on 12/10/2004 17:05 Ed Wilts said the following:

>On Tue, Oct 12, 2004 at 04:05:45PM +0100, Matthew Claridge wrote:
>  
>
>>I've got a basic FC2 installation, using software raid 1 across two IDE 
>>hard disks. Each hard disk is the master drive on its controller (so one 
>>is Primary master and the other is secondary master).
>>
>>If I remove the secondary master drive and boot the machine, all is 
>>well. If I remove the primary master, linux fails to boot (the bios 
>>cannot find an OS to boot). Even if I connect the secondary master hard 
>>disk to the primary master controller, it still cannot boot. So its as 
>>though linux hasn't mirrored the boot image successfully across the two 
>>disks.
>>    
>>
>
>Exactly.  There are 2 types of RAID 1 out there - one way is to mirror a
>physical volume (like VMS does) which mirrors everything.  The other way
>(like Linux does) is to mirror data partitions.  There are pros and cons
>to both, and you're getting hit by the con of the Linux approach.
> 
>  
>
>>There are three partitions on these disks, /dev/md[0-2]. /boot is on md0 
>>and / is on md1. Grub is installed on /dev/md0.
>>    
>>
>
>The boot block is what you're missing, and /boot isn't the same as the
>boot block.  The BIOS reads and executes the first physical sector of
>the chosen boot media on the systme.  Usually, this is contained in the
>first 512 bytes of the hard disk.  This is your 1st stage loader.  The
>boot loader is not part of any file system so software mirroring will
>not duplicate the data.  When you tell grub to install on /dev/md0, it
>doesn't actually write a boot block to both mirror members.
>
>  
>
>>There's an easy way around this, which is to use a boot disk, which 
>>should boot regardless of which disk has failed, and I know that linux 
>>will carry on working happily if either disk fails, but I'd like to find 
>>out why its misbehaving and correct it so that it can boot normally 
>>after a disk failure, making the whole thing a bit more resilient, 
>>especially if there's a delay obtaining a new hard disk or something.
>>    
>>
>
>What you need to do is duplicate the boot block.
>
>http://www.redhat.com/docs/manuals/enterprise/RHEL-3-Manual/ref-guide/s1-grub-installing.html
>
>        .../Ed
>  
>
that's pretty much as I thought. Tried the ideas on the link you gave, 
but this renders the system unbootable as it fails to load grub stage 2 :o(

maybe I'll just stick with a boot floppy......

thanks for the help though

Matt




More information about the redhat-list mailing list