Exabyte Hangs (LONG Digest of Responses)

[Alan J. Broder]

Dear fellow sun-spotters,

	Several weeks ago, I sent out the following message to sun-spots:

> We have an Exabyte tape drive attached to a Sun-3/180 running 4.0.1
> using the scsi-3 interface. We are having big problems. We backed up
> a whole disk two separate ways. First using tar, we did:
> 	tar cvbf 126 /dev/rst9 .
> then we did a dump of the same disk to ensure that we could
> recover our files. 
> 
> If you do a  
> 	tar tvbf 126 /dev/rst9
> tar proceeds to list all the files on the tape correctly. However,
> if you do a
> 	tar xvbf 126 /dev/rst9
> to restore all the files, tar makes it part of the way through
> and then just stops. At this point, tar can't be killed, even by
> root doing a "kill -9". The only way to ever use the tape
> drive again is to reboot the system. We discovered the exact
> same thing ocurring with the "dump" tape.
> 
etc.

Well, I received several responses which are digested below. I have
subsequently received several requests from sun-spotters to post
the results, so here goes...

As a result of the sun-spots responses I called sun and got the
patches which so far seem to have fixed the problem (these patches are
for Sun Bug Report ID 1012501).  Thanks to all for the help !

		Alan J Broder
		ajb at cornea.mitre.org


* ----------------
* 
* From: Randy Bruymbaugh (rando at skipper.dfrf.nasa.gov)
*
* We also have a 3/280 with an Exabyte using the supplied SCSI
* driver.  We have been doing mainly dumps.  I use a blocking
* factor of 1024 (instead of 126), not for any reason except that 
* when I tried different numbers, 1024 seemed to work.  We can dump
* once, but then dump hangs and can't be killed even by root 
* with kill -9.  
* 
* So... we called sun and they sent us a SCSI patch tape which
* fixed the hang problem.  Unfortunalely, now we can't use the
* Exabyte at all without getting DMA errors.  I have a call in to Sun
* now for this problem.
* 
* Most people will probably tell you to get the Sun 4.0.1 SCSI patch
* and it will fix your problems.  I would say that this is not
* really true with the Exabyte.  I'll let you know what Sun says 
* (when they call back).
* 
* ----------------
* 
* From: Bob Harris   (harris at basil-rathbone.mit.edu) +1 617 253 0725
* 
* We saw a similar problem right after we upgraded to 4.0.1.  (Our
* dumps to the Exabyte would hang mid-way through.)  The problem
* is with the scsi patched in 4.0.1.  Use /etc/remove_fiz scsi
* to back out the scsi "fixes" and run with the 4.0 version of
* the scsi drivers.
* 
* ----------------
* 
* From: Steve Hayman <sahayman at iuvax.cs.indiana.edu>
* 
* I called Sun today about another SCSI problem and they told me that
* they have a new SCSI patch ready that fixes some problems with
* scsi devices causing programs to hang.  They recommend de-installing
* the 4.0.1 SCSI patch and getting the new scsi patch tape.
* 
* ----------------
* 
* From: deltam!flyer!mark at uunet.UU.NET (mark galbraith)
* 
* Do you have SMD disks or SCSI disks?
* 
* If you only have SMD disks, you have run into a known bug in the
* SCSI driver.  Contact Sun for the fix.
* 
* If you have SCSI disks, then your problem might be new; but not 
* enough information was provided to analyze this over the net.
* Post more information to the net, and someone will probably be
* able to help (unless it is a hardware problem).
*
* ________________
* 
* From: Markus Diezi <diezi%ana.unibe.ch at RELAY.CS.NET>
* 
* We do have similar problems to read/write our 60MB tape cardridge
* which is connected by a Emulex-controller (SCSI). We could nor read
* sometimes the original distribution from sun.  After istalling the
* patch 4.0.2-SCSI, the problem was solved. Do you tried this already ?
* 
* ----------------
* 
* From: earle at Sun.COM (Greg Earle)
* 
* Known problem with a 4.0.1 machine having no SCSI disks but having
* a SCSI 1/4" tape drive.  If you are on Software Support, call the 800
* number and ask for the 4.0 SCSI patch tape to cure the tape drive hang problem.
* 
* ----------------
* 
* From: earle at Sun.COM (Greg Earle)
* 
* >My machine has an Exabyte SCSI drive, SCSI-3 interface, no other
* >SCSI devices.  And only SMD disks.  Does that fit the profile of
* >the known bug?
* 
* Other than the SCSI-3 host adapter (I don't think it is dependant on the
* type of adapter board used), this exactly fits the profile of the known
* bug.  It happens on systems with SMD disks, a 1/4" SCSI tape drive but no
* SCSI disks.  Which is what you have.
* 
* On the other hand, I have 2 SMD disks, a 1/4" SCSI tape drive *and* 1
* 327 Mbyte SCSI disk; I have *not* installed the patches and have not been
* bitten by the bug.  I have seen at least 3 customers at JPL get bitten by
* it, though.
* 
* ----------------
* 
* From: lxs at philabs.Philips.Com  (Ludhi Sidikpramana)
* 
* I have similar problems that you encountered, instead of problem
* reading the tape I had a problem dumping it ( SUN 4/280 ) but I have
* the exact problem on our VAX.
* 
* SUN:
* 	I called PerfectByte and they sent a scsi patch and since then
* the backup went fine except that when you try to use "mt -f /dev/nrst1
* status" it give you
* 
* 	...
* 	file no=0
* 
* The tape advanced alright, it just give the wrong info.	Any hints ???