Networker

Re: [Networker] R: [Networker] Red Hat Adaptec driver

2005-04-15 16:02:20
Subject: Re: [Networker] R: [Networker] Red Hat Adaptec driver
From: George Sinclair <George.Sinclair AT NOAA DOT GOV>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Fri, 15 Apr 2005 16:00:18 -0400
I would recommend calling LSI and checking with them. They do provide
drivers, but it might depend on your OS. For example, if you're running,
say RedHat Linux, then they may or may not have an RPM for your
particular kernel release. However, they should have the code so you
could build it into the kernel. Alternatively, the OS might just work
with it as is if the kernel already has support for it.

This particular card has two external VHDCI (Very High Density Connector
Interface) female connectors.  However, unlike the Adaptec cards, LSI
places channel A on the bottom and channel B on the top. No biggie, but
just something to keep in mind when you're hooking up cables. We're
running two libraries on the Dell Linux storage node: a Quantum P1000
with 2 SDLT gen 1. drives and a Storagetek L80 with 4 LTO Ultrium 1
drives. Both of these libraries have the standard 68 pin Micro-D female
connectors, so we needed Universal SCSI cables that had a 68-pin Micro-D
male connector on one end and a VHDCI male connector on the other.

Another thing I should note is that on both libraries, we do have the
picker daisy chained to drive 1 respectively. We didn't feel this was a
big deal since the traffic to the picker is minimal, so we saw no need
to put the picker on its own SCSI bus. However, on the Storagetek, we do
have drives 1 and 2 on their own SCSI bus, i.e. they are daisy chained
to the picker, and all three devices share the same SCSI cable to
channel A on the LSI host card. Drives 3-4 are on their own SCSI bus,
too, so they are daisy chained to each other, and both of these devices
share another SCSI cable to channel B on the host card. So, for the
Storagetek, we have two SCSI cables, each connected to a separate
channel on the host card. We figured putting each device on its own
channel was overkill, but having no more than 2 drives per channel
seemed wise. For the P1000, we did the same thing. This was probably
unnecessary since we only have one drive per channel, but we figured it
couldn't hurt. So, the P1000 uses a separate host SCSI card. This works
for us because the host has several different SCSI buses. Of course,
just because a host has say x number of SCSI slots doesn't mean it has x
number of separate buses since often times several slots may share the
same bus.

One thing I do see from time to time, though, is that we will get these
"Device or resource busy messages" on a device. It seems random and
appears to move around, never affecting the same drive in any
predictable manner. It occurs on both the libraries. We will also
sometimes see something like: "read open error. Device or resource
busy". I'm not sure what causes this. I've checked the NetWorker logs,
and I don't see anything suspicious other than the fact that it was
trying to load or unload a tape, but the tape will work fine later. It
seems that these messages eventually go away when the device gets used
again later. These "device or resource busy" messages, however, were
also occurring back when we were using the Adaptec cards, long before we
purchased the LSI cards. The only thing I know is that I rarely have to
reboot the host. Even when these messages occur, NetWorker is always
able to see the picker and all the devices unlike before when we were
using Adaptec cards. Back then, it seemed that often times NetWorker
would loose communication with the picker and sometimes one of the
devices. It also seemed that there was an unusually high coincidence
between these incidents and the "device or resource busy messages". 

I would be curious to learn if the LSI cards solve your problem, so if
you buy them, and they work, please post a follow-up. I think there may
very well have been a way to make the Adaptec cards play nice, but we
just could never figure it out. Another important thing to keep in mind,
too, is that because the Adaptec cards typically ship with the Dells,
you can be reasonably sure that they will be OEMed through Dell. So,
what happens if Adaptec issues a patch, but Dell doesn't have it yet?
They warn you about installing patches (flashing firmware) on an OEM
card if the patch is not from the same vendor and vice versa. As a
result, you might not be able to reliably patch the firmware on an OEM
card to make it work, unless Dell also has the patch, but what if they
don't? I mean, if you're using Dell RAID and you have OEM Adaptec cards
through Dell well ... natch, but when you're not using Dell
peripherals/equipment, and you're hanging it off a Dell OEM card then
... you see where I'm going. I wonder if a Dell tape library would have
worked? 

Sorry to be so long winded.

George

Librado Pamintuan wrote:
> 
> Hello George,
> We are planning to purchase this card, but we want to find ut first if it
> will work on our server and tape library.
> We have a Dell PowerEdge 2850 as the backup server and ADIC Scalar 100 tape
> library.
> Is the LSI Logic 22320-R card/adapter compatible with the server and tape
> library?
> 
> thanks in advance,
> Librado
> 
> On Fri, 8 Apr 2005 09:26:12 -0400, George Sinclair
> <George.Sinclair AT NOAA DOT GOV> wrote:
> 
> >We were having horrible SCSI bus reset problems with our Red Hat box and
> >Adaptec SCSI cards. I was literally rebooting that silly storage node
> >every day because when the problem would occur, the picker, and/or at
> >least one drive in the attached libraries, would simply disappear.
> >Nothing would bring them back except a reboot. you can only imagine the
> >frustration and utter chaos this was wreaking on our backups and tapes!
> >
> >We tried everything, and nothing worked. I don't think we ever tried
> >increasing the st buffers, though. Anyway, we finally gave up and just
> >bought some LSI Logic cards. The cards we purchased were the LSI
> >22320-R. This is an ultra 320 MB (160 per channel) dual SCSI SE/LVD
> >card. Presto! No more problems. In fact, the machine has been so stable,
> >it's unbelievable!
> >
> >The funny though is that we've had good success with Adaptec as far as
> >RAID, but very poor results with tape. I would be hesitant to ever use
> >Adaptec products for tape ever again, never mind the fact that Adaptec,
> >for whatever reasons, seems to have become the ubiquitous, omnipresent
> >SCSI card that comes defacto with just about anything you buy these
> >days. We just could never make it work. If it's that hard, why bother?
> >
> >George
> >
> 
> --
> Note: To sign off this list, send a "signoff networker" command via email
> to listserv AT listserv.temple DOT edu or visit the list's Web site at
> http://listserv.temple.edu/archives/networker.html where you can
> also view and post messages to the list. Questions regarding this list
> should be sent to stan AT temple DOT edu
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=

--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listserv.temple DOT edu or visit the list's Web site at
http://listserv.temple.edu/archives/networker.html where you can
also view and post messages to the list. Questions regarding this list
should be sent to stan AT temple DOT edu
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=

<Prev in Thread] Current Thread [Next in Thread>