From owner-freebsd-hackers Fri Jun 18 11: 7:45 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from dt054n86.san.rr.com (dt054n86.san.rr.com [24.30.152.134]) by hub.freebsd.org (Postfix) with ESMTP id A8C7914C9C for ; Fri, 18 Jun 1999 11:07:39 -0700 (PDT) (envelope-from Studded@gorean.org) Received: from localhost (doug@localhost) by dt054n86.san.rr.com (8.8.8/8.8.8) with ESMTP id LAA16203 for ; Fri, 18 Jun 1999 11:07:39 -0700 (PDT) (envelope-from Studded@gorean.org) Date: Fri, 18 Jun 1999 11:07:39 -0700 (PDT) From: Studded X-Sender: doug@dt054n86.san.rr.com To: freebsd-hackers@freebsd.org Subject: Re: Heavily loaded nfs/amd gets stuck Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG No action on this in -current for a few days, so let's try hackers. In response to some suggestions I tried raising the number of nfsiod's to 20 (the max) and increasing the sysctl cache value to 10, still no joy. I'm using amd to automount directories on sun (sol 2.6) server to my freebsd-current client machine. Details below, any help appreciated. Doug On Thu, 17 Jun 1999, Studded wrote: > > Also, should I be considering a move to -current for this box? Is > > -current stable enough right now to run a fairly heavily loaded web > > server? If the NFS in -current is going to be doing better than what's in > > -stable it will be worth a little headache to change, since our structure > > depends on it heavily. > > Well I went ahead and tried -current, and got better results, but > the same crash. With the following map: > > /defaults type:=nfs;opts:=rw,nosuid,vers=3,intr,proto=udp,noconn > * rhost:=IP${key};rfs:=/Space/${key} > > It went MUCH farther through the script (mounted about 60 out of 80 > directories) but it crashed just the same. Kernel stack trace looked like > this: > > stuff > Xresume1() > --- interrupt > bcmp() > mountnfs() > nfs_mount() > mount() > syscall() > Xint0x80syscall() Ok, another interesting development. What the script I'm running does is go through each user account on our sun servers, reads a file, then uses certain values from that file to print out conf files on the local freebsd server that's acting as an NFS client (and crashing). So it's mounting a directory, reading 250 files, mounting the next directory, reading the next 250 files, and so on for a total of 80 directories. I changed the script so that after each reading the 250 files for each directory it did a 'sleep 10' before it started again. This allowed the script to run through to completion. So, I'm still open to new things to try here. Does anyone have any suggestions? I've been looking at nfsiod, all I had started was the default 4 because I thought they would spawn more if they needed more, but apparently they don't. Would more of those help? Would turning them off altogether help? I *really* need help with this since my boss is (justifiably I think) loathe to put this box into service without a little more concrete evidence that NFS can hold up. Would it be better to send this to -hackers? Maybe file a PR? I don't mean to sound like a pest, and yes I know that we're all volunteers, etc. But after wheedling for 4 months to try freebsd I'm kind of feeling the pinch here. :-/ Thanks for any help you can provide, Doug To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message