[Linux-cluster] Re: GFS on md on shared disks?

[Jonathan E Brassow]

Hardware raid is fine, but software raid has it's limitations in this 
space.

If you can ensure that there will be no failures (in which case, you 
wouldn't be using raid5), and that the devices are always in the same 
order, etc -then go ahead.

If you have one machine that has a failure on a drive, only this 
machine will do the appropriate thing - leaving the others with a view 
of the device that is inconsistent with that of the machine on which 
the failure took place.  This is the fundamental problem.

  brassow
On Oct 7, 2004, at 11:07 AM, Ed L Cashin wrote:

> Erling Nygaard <nygaard at redhat.com> writes:
>
>> No, this will not work at all.
>>
>> All GFS locking is done on a filesystem level. In order to make this 
>> work
>> you need locking on the blocksystem level .
>
> I guess I'm looking for a concrete reason why it won't work.  I've
> been assuming it won't work, but I can't think of a concrete reason.
>
> You need to lock a resource group to allocate blocks for a file, and
> you need to lock the file in order to modify its blocks, so it's not
> entirely clear to me why you need block-level locking when md is
> involved.
>
> The only concrete problem I can think of is that the md has a
> superblock, and no node would know that other nodes are using the same
> md superblock.
>
> -- 
>   Ed L Cashin <ecashin at coraid.com>
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> http://www.redhat.com/mailman/listinfo/linux-cluster
>