From owner-freebsd-current Wed Sep 23 19:40:49 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id TAA07016 for freebsd-current-outgoing; Wed, 23 Sep 1998 19:40:49 -0700 (PDT) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from austin.polstra.com (austin.polstra.com [206.213.73.10]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id TAA06993 for ; Wed, 23 Sep 1998 19:40:45 -0700 (PDT) (envelope-from jdp@austin.polstra.com) Received: from austin.polstra.com (jdp@localhost) by austin.polstra.com (8.8.8/8.8.8) with ESMTP id TAA06964; Wed, 23 Sep 1998 19:40:24 -0700 (PDT) (envelope-from jdp) Message-Id: <199809240240.TAA06964@austin.polstra.com> To: Mike Smith cc: current@FreeBSD.ORG Subject: Re: shouting in a void? In-reply-to: Your message of "Wed, 23 Sep 1998 19:37:03 PDT." <199809240237.TAA05857@dingo.cdrom.com> Date: Wed, 23 Sep 1998 19:40:24 -0700 From: John Polstra Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > > I have a plan to eliminate the disk thrashing, but haven't > > implemented it yet. Once that's done, I believe it will truly be > > limited by the network link, unless you've got a T1 or better, or > > really slow hardware. > > Heh. You have to see what it does to a cvsup server to believe > it. 8) I have never seen a busy server (I'd like to sometime), but I do believe you. :-) It's actually a problem only for the master server, i.e., freefall. The mirrors now use "mirror mode" which doesn't have to do a tree walk for every client. Chris Timmons (cvsup.freebsd.org) tells me that it has made a tremendous difference. I could eliminate the tree walks on freefall by going to a periodic "cvsupscan" routine. The cost would be a delay in the visibility of file changes to the clients. When I've proposed it in the past, the universal response was no, we want updates from freefall to be up-to-the-second. I have another idea too, but it would require some help from cvs. I'm no longer adverse to that. > Even if you just cached the stat results against the mtime for > something in the repo, you'd win enormously. I'm not sure I understand what you mean. Please explain in more detail up in private mail. I'm _very_ interested in ways to reduce the disk load on the master server. John -- John Polstra jdp@polstra.com John D. Polstra & Co., Inc. Seattle, Washington USA "Self-knowledge is always bad news." -- John Barth To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message