From: "J. Bruce Fields" Subject: Re: Fixes for NFS in environments with large group memberships Date: Tue, 5 Apr 2011 19:35:52 -0400 Message-ID: <20110405233552.GB27961@fieldses.org> References: Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Cc: "linux-nfs@vger.kernel.org" To: "Finney, Sean" Return-path: Received: from fieldses.org ([174.143.236.118]:58282 "EHLO fieldses.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751859Ab1DEXfw convert rfc822-to-8bit (ORCPT ); Tue, 5 Apr 2011 19:35:52 -0400 In-Reply-To: Sender: linux-nfs-owner@vger.kernel.org List-ID: On Mon, Apr 04, 2011 at 10:02:59AM +0200, Finney, Sean wrote: > Hi There! >=20 > Last week I fired off a couple patches to this list and in hindsight = I > realize that I might not have been totally clear for the motivation > behind the patches: >=20 > [PATCH] Use a named constant for max number of managed groups, and in= c > [PATCH] Increase the default buffer size for cachelist channel files. I can't find those; could you resend? --b. >=20 > Since nobody has reviewed them yet, I thought I'd make it clear that > these changes (or something along such lines) are required for NFS > access to mountpoints when users are in largeish (>100) numbers of > groups. There are actually two problems: >=20 > * Hard-coded limit for group membership at 100 groups in mountd sourc= e > * Large server-side procfs writes are split into mulitple writes due = to > stdio buffering in mountd and svcgssd sources, resulting in client > hangs. >=20 > The patches in question solve this in as non-intrusive a manner as > possible, though I'm of course open to changing things up if you feel= it > should be done otherwise. So please let me know what you think, or i= f > it would be better for me to first go and open a bug, start a > discussion, etc. >=20 >=20 > Thanks, > Sean=20 > N?????r??y????b?X??=C7=A7v?^?)=DE=BA{.n?+????{???"??^n?r???z?=1A??h??= ???&??=1E?G???h?=03(?=E9=9A=8E?=DD=A2j"??=1A?=1Bm??????z?=DE=96???f???h= ???~?m