SCSI Tape problem with AHA1542

SCSI Tape problem with AHA1542

Post by Gruber Gerhar » Fri, 28 Nov 1997 04:00:00



I'm using an AHA1542 Adaptec SCSII controler with an Archive Anaconda
streamer tape attached. I tried to copy some data onto the tape (with
tar) and it worked fine. I could load it down again, append and
everyhting one expects from a tape. :)
The actuall problem I have is, when I put a bigger session on the tape
then mt (or the driver?) seems to hang after a while. I recorded about
300MB on the tape and it hung.
I can't kill mt and the tape is inaccessible (drive busy). After
rebooting the machine I tried to download all the data from the tape and
it worked again, but again it hung.

Before I started putting large sessions on the tape I tried it with
several smaller sessions (about 40MB each) and I had no problem at all.
Is there a limit to the maxmimum size of a single session? I also loaded
a tape with about 200MB on a tape with 150MB capacity to check if there
is a problem when the tape runs over, but this also worked. It seems as
if there is either a timeout due to the length of the session, or an
overflow. The only way to get the tape back working is to reboot the
machine.

Another question I have is: When the driver hangs is it possible to
reset the driver? I tried to install the driver as a module and unload
it after it hung, but this didn't work (as I expected).


you.

--

Bye,
   Gerhard





 
 
 

SCSI Tape problem with AHA1542

Post by R. Scott Baile » Tue, 09 Dec 1997 04:00:00



>I'm using an AHA1542 Adaptec SCSII controler with an Archive Anaconda
>streamer tape attached. I tried to copy some data onto the tape (with
>tar) and it worked fine. I could load it down again, append and
>everyhting one expects from a tape. :)
>The actuall problem I have is, when I put a bigger session on the tape
>then mt (or the driver?) seems to hang after a while. I recorded about
>300MB on the tape and it hung.
>I can't kill mt and the tape is inaccessible (drive busy). After
>rebooting the machine I tried to download all the data from the tape and
>it worked again, but again it hung.

No help here, but I'll report a similar problem, using an AHA1520 with an
attached Exabyte 8505XL 8mm drive (and a bunch of other stuff, including all
of this server's disks :-) We routinely backup the entire system, about 2GB,
and about 70% to 80% of the time everything is great. But it is not unusual
for the entire system to hang during this process (leaving activity light on
tape drive lit); it is necessary to hit the "reset" button to reboot the
system. For what it's worth, the severity of this problem seems to correlate
to the block size used during tape I/O; I use 16KB blocks now -- I could
hang the system nearly every time using larger block sizes (32K or 64K).
This behavior has been consistent from about 2.0.2x through 2.1.6x kernels.

If anybody has an answer or suggestion, please post -- I'd appreciate seeing
it.

Cheers,

    Scott Bailey


 
 
 

SCSI Tape problem with AHA1542

Post by Gruber Gerhar » Fri, 12 Dec 1997 04:00:00



> system. For what it's worth, the severity of this problem seems to correlate
> to the block size used during tape I/O; I use 16KB blocks now -- I could

I always thought that the blocksize depends on the tapedrive? But I'l
try it now.

Quote:> hang the system nearly every time using larger block sizes (32K or 64K).
> This behavior has been consistent from about 2.0.2x through 2.1.6x kernels.

Last time I used the tape I got my hang again. But this time the driver
reseted after I hit the eject button on the tapedrive and I didn't have
to reboot. If this works everytime, I'm "satisfied" but I'll try the
blocksize. Sombody mentioned also a new driver version but I don't have
the mail here. If you are interested in it, then send me a mail to

--

Bye,
   Gerhard





 
 
 

SCSI Tape problem with AHA1542

Post by Roland van Ho » Tue, 16 Dec 1997 04:00:00



:
: > system. For what it's worth, the severity of this problem
: seems to correlate
: > to the block size used during tape I/O; I use 16KB blocks now -- I could
:
: I always thought that the blocksize depends on the tapedrive? But I'l
: try it now.

Well I always use 10Kb blocks just like tar's default!

:
: > hang the system nearly every time using larger block sizes (32K or 64K).
: > This behavior has been consistent from about 2.0.2x through
: 2.1.6x kernels.
:
: Last time I used the tape I got my hang again. But this time the driver
: reseted after I hit the eject button on the tapedrive and I didn't have
: to reboot. If this works everytime, I'm "satisfied" but I'll try the
: blocksize. Sombody mentioned also a new driver version but I don't have
: the mail here. If you are interested in it, then send me a mail to

Yeah sometimes very rarely the process would hang and the button
would eject and the device was closed so my process would stop also.

My trouble is different, I use the 1542 with 3 HP DAT drives and
one 3c509 ethernetcard.
Occasionally I lose my 1542 connection and it's not resettable, this
is an known bug in the 1542 no solution.
But all was running 98% of the time well. Every week we shutted the
system down with powerdown so we wouldnot lose the 1542 in un unresettable
state.
More than a year everything went fine. We switched DDS drives now
one of the drives is DDS3. Everything went fine again.
Then I got a BAD idea, somebody suggested use 2.0.32.
So I did on another system. And it worked. So I thought ok I'll upgrade
my distribution as well and I'm using
a very old Debian distribution, the first that came out after 2.0.0
the boot floppies have 2.0.6 on them.
So I took the floppy with RH4.2 and reinstalled the system with already
2.0.32 on it, everything worked like a charm, faster and more functionality.
So I thougth I have to upgrade everything at least the distribution and
maybe also the kernel if the new functionality is desired (ISDN stuff).
The first try out was my backup machine since it's not used by anyone
but myself and can be rebooted as often as I like.
TROUBLES 2.0.32 doesnot keep the 1542 in the air for more than
20 minutes backup!
2.0.28 doesnot keep the 3c509 in the air for more than 3 hours
when backkingup 3 drives in parallel.
And it stays up for about 15 hours when backking up one tape
at a time.
I have a feeling it's the speed of the new compiler or utilities.
But I'm not in the position of debugging or analyzing it because
I don't know how and what to look for.
The machine is an old clone 486 33/66 Mhz, 32Mb ram 2 ide disks
about 2-3 Gb, a 3c509 and a 28k8 modem (not used during the backups
or backuptests).
When running the machine at 33 Mhz it stays up much longer
but the backup is so slow it's not an option, since I have to
at least backup the 3 tapes from 17:00 - 9:00 or 16:00 - 10:00
but I cannot use a bigger backup window than that.
Any ideas?

Roland

:
: --
:
: Bye,
:    Gerhard
:




--
Roland van Hout, software engineer at Foxim Software, Rotterdam,

 
 
 

SCSI Tape problem with AHA1542

Post by Gerhard Ahu » Fri, 19 Dec 1997 04:00:00



Quote:

>My trouble is different, I use the 1542 with 3 HP DAT drives and
>one 3c509 ethernetcard.
>Occasionally I lose my 1542 connection and it's not resettable, this
>is an known bug in the 1542 no solution.
>But all was running 98% of the time well. Every week we shutted the
>system down with powerdown so we wouldnot lose the 1542 in un unresettable
>state.
>More than a year everything went fine. We switched DDS drives now
>one of the drives is DDS3. Everything went fine again.
>Then I got a BAD idea, somebody suggested use 2.0.32.
>So I did on another system. And it worked. So I thought ok I'll upgrade
>my distribution as well and I'm using
>a very old Debian distribution, the first that came out after 2.0.0
>the boot floppies have 2.0.6 on them.
>So I took the floppy with RH4.2 and reinstalled the system with already
>2.0.32 on it, everything worked like a charm, faster and more functionality.
>So I thougth I have to upgrade everything at least the distribution and
>maybe also the kernel if the new functionality is desired (ISDN stuff).
>The first try out was my backup machine since it's not used by anyone
>but myself and can be rebooted as often as I like.
>TROUBLES 2.0.32 doesnot keep the 1542 in the air for more than
>20 minutes backup!
>2.0.28 doesnot keep the 3c509 in the air for more than 3 hours
>when backkingup 3 drives in parallel.
>And it stays up for about 15 hours when backking up one tape
>at a time.
>I have a feeling it's the speed of the new compiler or utilities.
>But I'm not in the position of debugging or analyzing it because
>I don't know how and what to look for.
>The machine is an old clone 486 33/66 Mhz, 32Mb ram 2 ide disks
>about 2-3 Gb, a 3c509 and a 28k8 modem (not used during the backups
>or backuptests).
>When running the machine at 33 Mhz it stays up much longer
>but the backup is so slow it's not an option, since I have to
>at least backup the 3 tapes from 17:00 - 9:00 or 16:00 - 10:00
>but I cannot use a bigger backup window than that.
>Any ideas?

What a horror story ! I'm using about the same configuration as you
are, but have no problems.

My setup:

486DX2/66
AHA 1542CF SCSI Controller
HP DAT drive DDS 1
Quantum Atlas 2Gb SCSI drive
SMC Elite Ultra Ethernet card
Kernel 2.0.30
Redhat 4.2
Maybee IDE has something to do with it ?

Gerhard.

--
Gerhard Ahuis          JO32EO               It's good to be independent

Unsolicited adverti*ts subject to $1000 consulting fee.