Skip site navigation (1)Skip section navigation (2)
Date:      	Fri, 9 Oct 1998 10:55:31 +0200 (METDST)
From:      Walter Hafner <hafner@informatik.tu-muenchen.de>
To:        freebsd-questions@FreeBSD.ORG
Subject:   strange page faults
Message-ID:  <13853.53123.507157.323111@hprbg5.informatik.tu-muenchen.de>

next in thread | raw e-mail | index | archive | help
Hi!

First of all: Sorry for the sparse information. But the machine in
question is not reachable at the moment (firewall). If needed, I can
provide more specific informations.

Here's the situation:

Pentium, 64 MB, ASUS Board, NCR-SCSI with 4 Disks from 1 to 4 GB each,
CD-ROM, DAT.

root, swap, usr on sd0, News spool on sd2 (a whole partition)

Until Sept. 98 the box ran 2.1.5, since then 2.2.7 with STABLE cvsup
We upgraded the system without changing the software in /usr/local.

Since yesterday we get strange page faults in innd. "strange" in the
context that we didn't change hard- or software since mid-september!

The box runs perfectly well as a server for a small ISP: apache, squid,
radiusd, msqld, ircd, rc5proxy, sendmail, ftpd and so on. The machine is
rather busy, but response time is ok and there are no problems.

The machine also runs a very ancient INN 1.4 feed. Incoming News are
delivered in UUCP-over-TCP batches every 30 minutes and as soon as rnews
starts working on the incoming batches, innd causes a page fault.

In the sync attempt, FreeBSD dies with one buffer still dirty. This is
also reproducable: It's always one buffer that can't be flushed. Note,
that this buffer is on different physical disks every time the machine
crashes.

Yesterday we changed all the RAM, temporarily disabled internal and
external processor cache and completely disabled paging on sd0. The page
faults still occur as soon as rnews starts.

Now I'm kind of stuck: It's not RAM, it's not disk. After the 2.1.5 -->
2.2.7 upgrade innd ran perfectly for 3 weeks.

What can be the reason? X-rays? Sun specles? El Nino? Spontaneous
mutations of innd? Or simply a code segment that wasn't executed since
the upgrade but was triggered by a specific batch?

Thanks.

-Walter

-- 
Walter Hafner_______________________________ hafner@in.tum.de
      <A href=http://www.in.tum.de/~hafner/>*CLICK*</A>;
 The best observation I can make is that the BSD Daemon logo
 is _much_ cooler than that Penguin :-)   (Donald Whiteside)

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?13853.53123.507157.323111>