From owner-freebsd-hackers Sun Jan 22 22:02:13 1995 Return-Path: hackers-owner Received: (from root@localhost) by freefall.cdrom.com (8.6.9/8.6.6) id WAA21566 for hackers-outgoing; Sun, 22 Jan 1995 22:02:13 -0800 Received: from brasil.moneng.mei.com (brasil.moneng.mei.com [151.186.20.4]) by freefall.cdrom.com (8.6.9/8.6.6) with SMTP id WAA21560 for ; Sun, 22 Jan 1995 22:02:11 -0800 Received: by brasil.moneng.mei.com (4.1/SMI-4.1) id AA13170; Sun, 22 Jan 95 23:59:45 CST From: Joe Greco Message-Id: <9501230559.AA13170@brasil.moneng.mei.com> Subject: No pages? revisited (2.0R) To: hackers@FreeBSD.org Date: Sun, 22 Jan 1995 23:59:44 -0600 (CST) Cc: ods-admin@ns.sol.net X-Mailer: ELM [version 2.4beta PL9] Content-Type: text Content-Length: 2377 Sender: hackers-owner@FreeBSD.org Precedence: bulk Odd 2.0R problems revisited...? I have been working with a large local BBS to get them Daemonized. I convinced them to provide FreeBSD shell accounts by offering to supply a router for their Internet connection (freeing up the machine they'd been using - a 4MB 386DX/40). I provided them with a Northgate 386DX/25 motherboard with 5MB last nite, and all seemed well. The hard disk, controller, serial card, display controller were all taken from the DX/40. The motherboard, memory, and WD8003E were taken from sol.net spare inventory and have worked in the past. It crashed early this morning with the dreaded "No pages" (it had been working just dandy on the 386DX/40). They rebooted. It crashed again tonite. The machine is not doing anything at the time of crash, at least that I am aware of - other than acting as internetwork router. There is a chance that it had started /etc/daily this morning, but reports were that it crashed prior to that. Nobody logged in. Just running pppd, named, and trivial other standard stuff. Notes: I did notice at one point that the machine seemed to be mildly funky - it got in a state last night where it was working, but systat would report "Alternate system clock has died, reverting to pigs" on some screens, and vmstat's output was rather funky. We let it go, since it seemed otherwise happy. Systat/etc works now, after a power cycle. Also: the kernel was generated from a 2.0R source tree with a single patch previously supplied to me by David Greenman(I think) that tweaked some magic number in vm/vm_pageout.c from 4 to 16 (vm_pageout_free_min = 16 + something) because I had had a similar problem on my news server (I am not sure if the extra memory I added or the patch fixed it on my news server, but I haven't seen it happen again). So this really makes No Friggin' Sense. As an interim fix, I am going to put 8MB more in the machine tomorrow morning. I suspect some sort of hardware problem - because it WAS working great on their DX/40 - so I may end up swapping in a new motherboard. I'd hate to do that, because this is a great "New Life for Old Hardware." Comments/advice/etc welcome. ... Joe ------------------------------------------------------------------------------- Joe Greco - Systems Administrator jgreco@ns.sol.net Solaria Public Access UNIX - Milwaukee, WI 414/342-4847