Date: Mon, 18 Dec 2000 10:00:19 -0500 From: "Salyzyn, Mark" <mark_salyzyn@adaptec.com> To: "'Mike Smith'" <msmith@freebsd.org>, Micah Anderson <micah@indymedia.org> Cc: noah <noah@indymedia.org>, freebsd-scsi@freebsd.org Subject: RE: update Message-ID: <50DB155AD0CED411988E009027D61DB31D8116@otcexc01.otc.adaptec.com>
next in thread | raw e-mail | index | archive | help
Although I figure Adaptec's Tech Support would be the best to know about
generic issues with drive access, the possibilities for this issue could be:
1) No cable and/or drive cabinet domain validation, so one might have to
roll the SCSI speed down a bit to compensate for cable and/or drive
combination issues.
2) Some drives are more comfortable with either over (more than just the two
endpoints) or under (only the last drive or controller) termination.
3) Contact tech support for a later Firmware release, there may be known
issues with your drives, cabinets and/or drive combinations that might have
been addressed with either drive firmware, or controller firmware updates.
Currently the customer has better access to Technical Support than I do at
this moment :-( even though I virtually end up driving over top them each
morning as I head to the parking lot ...
In any case, I will report this to the Firmware engineers to see if they
have any additional comments to add about this issue.
Keep in mind that at initial negotiation, the speed is lower, the transfers
less stressful, than at operating system time. Edge issues may surface as a
result, sometimes appearing different from OS to OS. For instance, I believe
the ASR driver can request up to 58 (~4KB) scatter/gather elements in one
request, allowing up to 256 requests/device. NT's scsiport driver, on the
other hand, limits request to 64KB/each and only 16 requests/controller.
Stresses vary.
However, OS issues do not typically affect drive failures, which is curious.
I have an issue that comes up in FreeBSD, for instance, with the array
performance in an impacted (read failures do not fail an array since data
can be reconstructed) state since the requests take much longer to fulfill
than in the genuine failed state. Impacted means every request still tries
to be fulfilled by first trying to talk to the not-yet failed component.
This has the catch-22 effect of not being able to mount the array head due
to the protracted responses on some failed drive scenarios before the
adapter has considered the component to be marked as failed. Pulling the
errant drive might be the only way. Later adapter Firmware may deal with
this through careful consideration of request response time. Tech support
may supply a select fail-on-read firmware/NVRAM, or one can chose to bump up
the timeout in the SCSI layer. This issue, for instance, does not occur
under Solaris because their SCSI layer is set to 2 minute timeouts.
Sincerely -- Mark Salyzyn
-----Original Message-----
From: Mike Smith [mailto:msmith@freebsd.org]
Sent: Monday, December 18, 2000 5:37 AM
To: Micah Anderson
Cc: noah; freebsd-scsi@freebsd.org; mark_salyzyn@adaptec.com
Subject: Re: update
Mark; I miscopied you on my previous reply to this message, sorry about
that. Do you have any ideas?
> On Sat, 16 Dec 2000, Mike Smith wrote:
>
> > > At "asr0: major=154" the raid card begins a high pitched beep
indicating
> > > that two of the drives have failed and that a rebuild of the raid is
> > > required, but we've tested all of the drives and replaced the raid
card
> > > with a new one, and still get the same problem. The reason I'm asking
> > > about possible software issues is that other OS's have worked on this
raid
> > > setup.
> >
> > I've copied Mark at Adaptec, who is the author and principle maintainer
> > of the 'asr' driver, since he's going to have the best idea of what's
> > actually going on here. Without saying which OS' you've used, it's
tough
> > to know whether they simply aren't enabling the card alarm though.
>
> We have gone through exhaustive troubleshooting lengths to try to
determine
> what the problem is. I have swapped RAID cards, swapped cables, tried a
> different motherboard, different powersupply in every possible combination
> of configuration. Each time I have to start from the beginning, destroying
> the RAID configuration and then creating a new one, which takes over an
> hour, so this process has taken literally three weeks to try all the
> potential configurations.
>
> The RAID alarm goes off on the card during the FreeBSD boot process, the
OS
> continues to boot, but the alarm continues. Rebooting and going into the
> Adaptec setup tells us that a drive has failed, it is not the same drive
> every time. During bootup after the RAID POST when the SMOR utility is
> loading it will usually show the RAID-5 drive as well as the single drive.
> It is almost as if one of the drives of the RAID is pushed out of the
RAID.
> Individually, each drive works fine. If I install FreeBSD on a single
drive,
> without a RAID constructed things act as normal. These are IBM 10k RPM
LVD
> drives and I ran IBM's drive test utility on each one of them and it came
> back with no errors.
>
> I have been able to install Debian Linux and use the card/drives without
> this problem. I have called Adaptec to ask them about this and was told to
> try changing the drive speed from Ultra 3 to Ultra as well as change the
> delay from the default to 30 seconds, all of these do not change the
> behavior whatsoever.
>
> I have spoken with one other person who had a similar type of problem,
> except what was happening to him was he was loading some DOS drivers, one
of
> which would wipe the RAID card configuration when it was loaded (ASAPI? I
> can't recall right now)... I am wondering if there are some other drivers
> that are being probed in the generic FreeBSD kernel that are doing a
similar
> thing to the config.
>
> >
> > Have you tried running the Adaptec management software to check the
> > status of the card?
>
> In FreeBSD? If there is such a thing it would be interesting to know where
> one could get it. The CD that was included with the card has no FreeBSD
> anything on it - the website has no FreeBSD information or downloads on it
> (except for the breif mention that it is supported, but if you call for
> support you can't get it). Or are you talking about the SMOR utility that
> you can access from the BIOS?
>
> Thanks for any help that you can offer.
>
> Micah
>
>
> >
> > > Thanks.
> > >
> > > On 12/15, Mike Smith wrote:
> > > >
> > > > > Hi, I'm working on trying to install FreeBSD 4.2 on a dual p3 700
with
> > > > > an Adaptec 3200S raid card. From what I can tell everyone that has
tried
> > > > > this card has had good luck. When we install FreeBSD (booting off
cd) it
> > > > > recognizes the card and installs on it perfectly, but when it
loads the OS
> > > > > off the raid it does something to damage the hardware raid,
requiring us
> > > > > to rebuild the RAID in the 3200S' bios. We're pretty sure that
this isn't
> > > > > a hardware problem.
> > > >
> > > > You haven't actually included anything that suggests that there's a
> > > > problem occurring, so it's somewhat difficult to guess what's going
on.
> > > >
> > > > However, I don't lend much credibility to the suggestion that
"FreeBSD
> > > > does something to damage the hadware raid" - things just don't
happen
> > > > like that.
> > > >
> > > > I would be inclined to suspect that you probably have a suspect
disk, or
> > > > cabling/enclosure problems, but without more details it's hard to be
sure.
> > > >
> > > >
> > > > --
> > > > ... every activity meets with opposition, everyone who acts has his
> > > > rivals and unfortunately opponents also. But not because people
want
> > > > to be opponents, rather because the tasks and relationships force
> > > > people to take different points of view. [Dr. Fritz Todt]
> > > > V I C T O R Y N O T V E N G E A N C E
> > > >
> > > >
> > >
> > > --
> > > noah .. email for pgp/gpg key
> > >
> >
> > --
> > ... every activity meets with opposition, everyone who acts has his
> > rivals and unfortunately opponents also. But not because people want
> > to be opponents, rather because the tasks and relationships force
> > people to take different points of view. [Dr. Fritz Todt]
> > V I C T O R Y N O T V E N G E A N C E
> >
> >
>
--
... every activity meets with opposition, everyone who acts has his
rivals and unfortunately opponents also. But not because people want
to be opponents, rather because the tasks and relationships force
people to take different points of view. [Dr. Fritz Todt]
V I C T O R Y N O T V E N G E A N C E
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?50DB155AD0CED411988E009027D61DB31D8116>