Subject: Latest problems
To: None <current-users@sun-lamp.cs.berkeley.edu>
From: Robert L. Shady <rls@zeus.id.net>
List: current-users
Date: 05/04/1994 08:17:01
Couple quick questions/comments..
I am running currents from the past 2 weeks or some (Apr 20+) on the
following configurations
ZEUS HADES GATEWAY1
==== ===== =======
486/DX40 VLB 386/DX33 386/DX33
16MB RAM 8MB RAM 4MB RAM
Adaptec 1542C Generic IDE Cntrlr Generic IDE Cntrlr
180MB SCSI-2 340MB IDE 40MB IDE
1GB SCSI-2 540MB IDE
250MB Wangtek Soundblaster PRO
NE-2000 Clone NE-2000 Clone NE-2000 Clone
SVGA Card/Monitor SVGA Card/Monitor Mono Card/Monitor
4 16550 Serials 4 16550 Serials
2 28.8k Modems 2 28.8k Modem
2 14.4k Modems 2 14.4k Modems
Gateway1 is our SLIP connection to the Internet. It is also our nameserver,
and dialup SLIP/PPP server. Hades is our news server. Zeus is our main
Interactive dialup server, and backup nameserver.
1) All machines exhibit some form of crash/reboot at least once every 3 days
or so. Gateway1 reboots SEVERAL times a day, generally three or four
times per hour under normal->heavy load (Average 1000+ cps constant).
Is there still a problem with only 4 MB of ram? It seems as thought there
is some form of major problem with tcp/ip code that is causing a crash.
How would I go about tracing this more closely (assuming I added the
DIAGNOSTIC & DDB options to the kernel, how do I trace the stack/etc).
2) What is the best way to avoid an NFS server 'hang'. I am cross mounting
several directories between Hades & Zeus, and when one or the other is
down, the other hangs indefinately until the other one comes back up.
I am using the following in my /etc/fstab:
/dev/sd0a / ufs rw 1 1
/dev/sd0e /tmp ufs rw 1 2
/dev/sd1a /home ufs rw 1 3
/dev/mcd0d /cdrom isofs ro 1 3
kernfs /kern kernfs rw 1 1
procfs /proc procfs rw 1 1
hades:/ /hades nfs rw,soft,intr,bg 1 3
hades:/news /news nfs rw,soft,intr,bg 1 3
3) I am still getting a massive amount of SILO overflow's on Gateway1, even
though I am using the 16550 UARTS. This is notices with sustained
transfer rates of ~1700cps.
4) Was there a good reason for removing/changing the definitions of some of
the most common things (ie: sys_errlist, etc)? I asked this before, and
I believe that the answer was to become POSIX compatible, but this is
starting to get rediculous. I have to really tear apart programs to get
them to compile under NetBSD now, where as before, if they worked on a
Sun, they probably compiled under NetBSD without change. Is there some
way we can put the old defines back, maybe under a "#ifdef PRE_POSIX"
or something to make it easy to port old code that may not yet conform?
-- Thanks!
------------------------------------------------------------------------------