Nesting gvinum?

[Jonathan Dobbie]

Rick C. Petty wrote:
> On Wed, May 09, 2007 at 09:43:14AM -0500, Jonathan Dobbie wrote:
>   
>> I was hoping to have something like md+LVM where I could change the size
>> of RAID 5 volumes.  Could I have a large gvinum raid-5 volume and then
>> make 3 gvinum volumes on top that could be resized?
>>     
>
> Except for the nesting, this is precisely what gvinum does.  Why do you
> need nesting?  Resizing volumes is something gvinum is supposed to be able
> to do.
>
>   
I thought that it could only resize striped volumes?  Would you happen
to have a guide handy?  If I, say, wanted to resize usr and log to add
20GB to log, could I drop the s0 subdisk from both, resize, readd, sync
and then repeat for s1 and s2?

>> Also, is there a better fs to use on top of gvinum?  Unlinks and file
>> creation "feel" slow to me.  I thought about reiser, but my coworker was
>>     
>
> UFS2.  It's incredibly fast, even with RAID5 underneath.
>
>   
I used newfs and sysinstall to create all of the partitions, so they
should be UFS2, no?
>> burned the last time he put it on a production box and we've been rather
>> cautious since.
>>     
>
> I wasn't aware that FreeBSD supported reiser.
>
>   
I was a but surprised as well.

*HISTORY* <http://www.freebsd.org/cgi/man.cgi?query=mount_reiserfs&apropos=0&sektion=0&manpath=FreeBSD+6.2-RELEASE&format=html#end>
     The *mount*/_/*reiserfs* utility first appeared in FreeBSD 6.0.

*AUTHORS* <http://www.freebsd.org/cgi/man.cgi?query=mount_reiserfs&apropos=0&sektion=0&manpath=FreeBSD+6.2-RELEASE&format=html#end>

     The *mount*/_/*reiserfs* utility and manual were written by Jean-Sébastien
     Pédron <dumbbell at FreeBSD.org <mailto:dumbbell at FreeBSD.org>>.

>> Didn't UFS2 come out for freeBSD a while ago?  I'm
>>     
>
> as of 5.0, in the 2002-2003 timeframe.
>
>   
>> I know that shelling out the 250$ for a raid card is the best option,
>> but no one with a budget wants to put any money into the box.
>>     
>
> Not really.  Software RAID has been proven to be faster than hardware
> numerous times.  I read somewhere that this has to do with the transfer
> blocksize or something.  I haven't witnessed this myself, but I've seen
> gstripe & gmirror outperform RAID0/1 cards.
>
>
>   
I'm definitely sold on software RAID for 0 and 1, but it is nice to boot
off of RAID 5 and forget that it is there until a drive dies.