Subject: kern/18636: Multiple uvm_pagefaults
To: None <gnats-bugs@gnats.netbsd.org>
From: None <don+send-pr@resun.com>
List: netbsd-bugs
Date: 10/12/2002 17:47:49
>Number:         18636
>Category:       kern
>Synopsis:       Multiple uvm_pagefaults
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Oct 12 17:49:00 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Don Phillips
>Release:        NetBSD 1.5.1
>Organization:
  Don Phillips         don@resun.com
  Escondido, Calif.    My opinions are just that, and no more.
>Environment:
	
System: NetBSD pulsar.resun.com 1.5.2 NetBSD 1.5.2 (resun) #1: Fri Aug 16 01:18:37 PDT 2002 don@pulsar.resun.com:/usr/src/sys/arch/i386/compile/resun i386


>Description:
        When running a SW system that uses Perl 5.6 & Perl 5.0, two mysql DBs,
        one of which has tables that have grown to over 1.2GB, the system
        becomes unstable.  The same system had been running for 18 months w/o
        problems.

        The HW was completely rebuilt, using different MB manufacturers.  New
        memory modules were used.  Some networking cards were shared.  There is
        a tlp onboard ethernet.

        The system would sometime hang, w/o the ability to get a crash dump.
        Sometimes, the dumpdev would be reported as not ready.
>How-To-Repeat:
	I suspect that this is specific to my configuration, as there have
        been no other reports of the problem.  If I reenable the SW subsystem,
        the system generally won't survive for more than an hour before
        crashing.
>Fix:
	Not running the SW subsystem has worked for a month.  :-(

        I've tried migrating to 1.6, and have experienced similar problems,
        without running the SW system.  I can't even get a one-time-through
        compile of a kernel, without having segfaults or other problems in the
        compilation.  The compilation will eventually produce a working kernel,
        if I push through the errors.
>Release-Note:
>Audit-Trail:
>Unformatted: