🚀 go-pugleaf

Don Smith wrote:
> 
> I have a single server 4.1 network, pentium processor, 32
> megs of ram, 1.2 gig SCSI hard drive. SCSI card is Adaptec
> 2940. Also have a Conner tape drive and a Plextor 4X CD on
> the SCSI buss, but have disconnected them (except when I
> need to run a tape backup. See next paragraph)
> 
> The server will run fine for about as long as I leave it
> alone. Because of problems we are having we have disabled
> automatic tape backup. However, if I go in and run a tape
> backup (Backup Exec 7.0, Conner SCSI tape drive), the next
> morning the server will lock up completely for anywhere from
> 3 to 7 times. Each time, the only solution is to power it
> down and bring it back up. At that point, we usually down it
> gracefully, but it may still lock up for several more times.
> 
> Question: What is different about the morning after I run a
> tape backup? The problem happens even if I down the server,
> and disconnect the tape drive after running the backup.
> Could it be that either running the tape backup or downing
> the server is causing Novell to compress a large betrieve
> data file (4+ megs) which is from a very old version of
> betrieve. Then the following morning, when someone uses that
> file, something is causing the SCSI bus to hang?
> 
> I realize I have to find the cause of the bus hanging. But I
> am totally at a loss as to why the server will stay up fine,
> apparently indefinitely, after it finally gets over its
> problems, but will fail 100% of the time the morning after I
> run a tape backup.
> 
> Thanks for any ideas.

Your problem is most likely the crappy nlms that run your tape backup 
program. If this program doesn't use any nlms it could also be support 
nlms loaded when the program loads. Your solution: in Netware 4.x you can 
load the domain.nlm (load first thing in startup.ncf, you'll have to 
reboot once for this, but not every time) then switch to the OS_PROTECTED 
domain, I believe the command is "SET DOMAIN = OS_PROTECTED" at the 
console prompt, now every nlm that is loaded will run in Intel protected 
mode, whereby these nlms will NOT crash your system. The only tradeoff is 
a little speed loss on THOSE nlms which are loaded there. Now, when they 
crash you can read the console, or set conlog working and you'll know 
what happens when those nlms crash themselves. (check my use of the 
domain nlm, as I'm going by memory here).