>Number:         7620
>Category:       kern
>Synopsis:       kernel panic (machine self reset)
>Confidential:   no
>Severity:       critical
>Priority:       low
>Responsible:    kern-bug-people (Kernel Bug People)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri May 21 15:20:01 1999
>Last-Modified:
>Originator:     Brad Knotwell
>Organization:
>Release:        1.4
>Environment:
NetBSD1.4 i386
>Description:
under moderate load (thttpd@130 connections/s sustained throughput),
the box will reset itself within about a 5 minute period.  only a
couple of times did I drop in to ddb, the stack trace was
       _gdt_put_slot
       _tss_free
       _switch_exit
       _bpendtsleep
       _uvm_scheduler
       _main

NOTE:  this box was a backend behind a BSDI3.1 based load balancer. The
client was a Linux2.2.5 machine.  The balancer basically requested the
same page over and over.

Other factors:  the number of sockets in a TIME_WAIT state was pretty
high.  Similarly, the size of the pagedaemon,reaper,and swapper
grew to be large as well (at least in several of the cases). netstat
-m didn't provide any clues.

if you need more information, drop me a note and I'll run the test
again.
>How-To-Repeat:
start a traffic generator against a NetBSD1.4 box and let it run
sustained for a short period of time.
>Fix:

>Audit-Trail:
>Unformatted: