On 2006/09/23, at 11:30, Claude Jones wrote:
On Friday September 22 2006 8:52 am, Joel Rees wrote:
I don't like the idea, because I am aware that many ATA controllers
can't seem to keep master and slave out of each other's hair.
I have to ask - several times in this post you allude to master/
slave issues.
What are you talking about?
Good question. Like I said in my earlier post, the advice my be
outdated. For all I know it may not apply to Linux kernel and drivers
written after a certain year, probably before 1999 if that's the
case. I've never seen it myself that I know of, but then I've mostly
avoided MSWindows compatible hardware. And the companies I've worked
for have tended to not fully populate a channel.
You could plug something like "master slave ata corruption" in a
search at Google and get some interesting reading, though.
The sales guy at Pasokon Kobo over in Mikage a couple of years back
warned me against trying to use both master and slave at the same
time, but they admitted to not having much experience with Linux. In
the MSWindows universe they were familiar with, there were enough
reports of silent corruption they figured it was best to stay with
one channel, one drive, particularly when building RAID on ATA
controllers.
Correct me if I'm wrong, but master and slave on a single channel
can't handle commands concurrently? What I recall reading was that it
was known that some companies' drivers for MSWindows were known to
fail to keep state straight between calls, but I remember reading
some guy who said that even the hardware would lose state in certain
situations. And the corruption would tend to be silent, no warnings
until it was too late.
I have been building my own machines since 1989,
and I've managed hundreds in my work. Nearly all the machines I've
worked on
had dual drives with master/slave configurations - I use computers
in pretty
stressful environments such as video and audio editing workstations
a lot.
MSWindows or Linux? BSD? Mac? Controllers limited to certain
manufacturers? Did the usage patterns tend to be such that only one
drive was used at a time, as in boot and run this drive or that, only
use the other drive for backups performed during break? Can you
guarantee there was no silent corruption? (regular diffs between
backups or such?)
Yet, in all that time, with all those machines, I've never encountered
problems with drive conflicts, except ones I've caused myself due
to improper
jumpering. For the second time in a month on this list, someone has
alluded
to such problems but with no details. What sorts of problems have
you had? I
would like to know in case I encounter such in the future.
Like I say, I have not seen drive failures I can pin on this
particular issue. What I know is second hand, from the web. I have
seen drives fail much earlier than was reasonable, but the failure
patterns seemed to be more related to the issue of ATA drives failing
to sync the cache to the disk when told to, or maybe even just cheap
recording surfaces.
