gmirror and partitioning

[Gayn Winters]

> From: Nagy László Zsolt [mailto:nagylzs at freemail.hu] 
> Sent: Wednesday, May 17, 2006 12:29 AM
> To: gayn.winters at bristolsystems.com; freebsd-questions at freebsd.org
> Subject: Re: gmirror and partitioning
> 
> 
> 
> >> acd0: CDROM <GCR-8523B/1.01> at ata0-master PIO4
> >> ad8: 152626MB <SAMSUNG HD160JJ ZM100-33> at ata4-master SATA150
> >> ad10: 152627MB <SAMSUNG HD160JJ ZM100-33> at ata5-master SATA150
> >>     
> > Sadly, yes; or buy a bigger second disk. You could do the 
> install on the
> > smaller disk first, or you could make ad10s1 smaller.
> >   
> Okay, I did this. But I'm still interested in the topic. The 
> size of a 
> hard drive is determined by the manufacturer. It depends on how many 
> sectors, heads and cylinders present in the device. The 
> actual available 
> size can be smaller because of bad sectors on the disc. But 
> the BIOS (or 
> FreeBSD) should detect the full size, including all sectors. These 
> devices are identical. Then how in the hell could it add one 
> more MB to 
> the second device? I presume if I swap drives between ata4-master and 
> ata5-master then still ad10 would be bigger. Is this a bug in FreeBSD?
> 
>    Laszlo

Hi Laszlo,

No bugs; I think this is normal.  Both the BIOS and the OS are only
going to see the blocks the hard drive thinks are useable.  Even though
the drive geometry is "fictitious", most people still recommend defining
your slices to begin and end on cylinder boundaries.  I assume this
makes accessing blocks in the slice a few nanoseconds faster, but I'm
actually not sure.  Perhaps someone on the list knows more detail.  In
any case, if you do this for as10s1 gmirror will replicate this property
to the other disk.  You'll also most likely have a few blocks left over.
As for units of MB, I'm not sure.  Could be rounding.

Can we assume your gmirror is now working?

Best regards,

-gayn

Bristol Systems Inc.
714/532-6776
www.bristolsystems.com