Return-Path: Received: from mx2.netapp.com ([216.240.18.37]:27891 "EHLO mx2.netapp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752988Ab1F0SmH convert rfc822-to-8bit (ORCPT ); Mon, 27 Jun 2011 14:42:07 -0400 Subject: Re: Mounstats keying on fsid or ipaddr? From: Trond Myklebust To: Joe Frost Cc: linux-nfs@vger.kernel.org Date: Mon, 27 Jun 2011 14:41:36 -0400 In-Reply-To: <4E08C8BE.2000407@pixar.com> References: <4E08C8BE.2000407@pixar.com> Content-Type: text/plain; charset="UTF-8" Message-ID: <1309200096.25643.10.camel@lade.trondhjem.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On Mon, 2011-06-27 at 11:15 -0700, Joe Frost wrote: > We've been using mountstats to graph/track NFS usage based on path. > We've recently made some changes on how we mount filesystems here where > it's possible the same IP will be used for multiple mounts. Since that > we've noticed that all mounts to the same IPaddr are showing the same > statistics. Wondering if anyone had ideas why it might be keying off of > IPaddr? mountstats keys off the fsid as reported on the _server_. If your server is reporting that these directories are all contained in one big volume, then that will suffice to confuse the client. One way to work around this, is to tell the client that the subdirectories are on different volumes. You can do that using the "nosharecache" mount option. Beware, however, that if you have hard linked files that appear under more than one of these mountpoints, then the client will treat them as if they are completely separate files for the purposes of caching and close-to-open cache consistency checking. Cheers Trond -- Trond Myklebust Linux NFS client maintainer NetApp Trond.Myklebust@netapp.com www.netapp.com