[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: SCSI Device Addresses


On Sat, 2008-03-01 at 22:31 +1100, Hamish Moffatt wrote:
> On Sat, Mar 01, 2008 at 11:00:01AM +0000, Steve Dobson wrote:
> > On Sat, 2008-03-01 at 14:31 +1100, Hamish Moffatt wrote:
> > > Doesn't udev assign constant sdN/hdN identifiers? It appears to look up
> > > the drive serial number and attempt to do this.
> > 
> > From what I've observed hdN is consistant.  hda is *always*
> > primary-master,
> > hdd is secondary-slave, and so forth.
> Actually I don't think that is true with our udev configuration. It
> tries to keep the same ID assigned to the same physical disk, so if you
> rearrange them they maintain their IDs, and hence your fstab still
> works.

As I said this behaviour is what I have *observed*.  I'm pretty sure
that on a system a while back were I had a SATA boot disk and a
CDROM/DVD driver plugged into the secondary IDE it was there as hdc.  

And I think that many moons ago I moved a CDROM drive from the primary
to the secondary IDE and I got hdd without a hdc, but that was so long
ago my memory maybe playing tricks.

The behaviour maybe IDE driver specific and therefore dependant on the
IDE chips used on the motherboard, I don't know.  I do have a tendency
to build my systems consistently, CDROMs only go on the primary IDE bus
if there are not PATA drives in that system, and most of the time there
have been, so I have found that the drive letters for hd devices are
where I expect them to be.

> > > Also, if you run LVM or software RAID then you get static identifiers
> > > out of those automatically, even with the underlying disks moving
> > > around.
> > 
> > Yes, but you need fixed physical device addresses to create the LVM or
> > software RAID devices in the first place.
> No you don't. LVM puts unique identifiers on each physical volume, and
> will scan all disks to find them. It doesn't hardcode their locations
> anywhere. I think software RAID can also do this.

I didn't know that, thanks for the heads-up.


Reply to: