[Beowulf] Software RAID?

[Mark Hahn]

>> First:  No hot swap.  You can do a kind-of-cold swap (have to take the
>> mount offline, and can execute a few MD raw-disassemble, and then turn
>> the device off, swap, force Linux to rescan the scsi bus, mark the drive
>> as a hot spare, and force reassembly ... then remount).  This may or may

I think some of this is obsolete or at least easier ways exist now.
on the other hand, the MD maintainer is quite conservative when it 
comes to trusting MD to "do the right thing automatically", since 
such policies seem to always eventually bite someone...

>> not work, depending upon the linux driver for the SATA port.  Some get
>> very unhappy if the drive goes away after it found it.
> Is this primarily a driver issue? Are there any provisions in libata for
> rectifying it (if it's a software problem only)?

http://linux-ata.org/software-status.html#hotplug

I think the sata phy is a bit network-like, in the sense that the controller
can detect when a live device is on a port.  I haven't experimented with
whether this awareness is integrated into the udev/hotplug framework, 
and whether udev already has scripts that might auto-do-the-right-thing.

auto-DTRT is easier in "enterprise" environments, since, for instance, 
the replacement disks in my HP MSA20 arrays are all pre-initialized by HP.
by contrast, if MD tried to auto-DTRT, it would have to cope with people 
jamming in random disk with random contents.  consider, what might happen if
you stick in a replacement disk which happens to claim to already be part of
a live raid _and_ that its generation number is higher...