From: Charles-Edouard Ruault Subject: Re: Re: Linux 2.4.25, nfs client hangs when talking to a MacOS nfs server. Date: Thu, 18 Mar 2004 17:53:13 +0100 Sender: nfs-admin@lists.sourceforge.net Message-ID: References: <482A3FA0050D21419C269D13989C61130435DD8A@lavender-fe.eng.netapp.com> Mime-Version: 1.0 (Apple Message framework v613) Content-Type: text/plain; charset=US-ASCII; format=flowed Cc: "" , "" Return-path: Received: from sc8-sf-mx2-b.sourceforge.net ([10.3.1.12] helo=sc8-sf-mx2.sourceforge.net) by sc8-sf-list2.sourceforge.net with esmtp (Exim 4.30) id 1B40mR-000558-Lt for nfs@lists.sourceforge.net; Thu, 18 Mar 2004 08:54:03 -0800 Received: from czame.czame.com ([64.33.120.2]) by sc8-sf-mx2.sourceforge.net with esmtp (TLSv1:DES-CBC3-SHA:168) (Exim 4.30) id 1B40mM-0006LC-1f for nfs@lists.sourceforge.net; Thu, 18 Mar 2004 08:53:58 -0800 In-Reply-To: <482A3FA0050D21419C269D13989C61130435DD8A@lavender-fe.eng.netapp.com> To: Trond Myklebust , "Lever, Charles" Errors-To: nfs-admin@lists.sourceforge.net List-Unsubscribe: , List-Id: Discussion of NFS under Linux development, interoperability, and testing. List-Post: List-Help: List-Subscribe: , List-Archive: On Mar 17, 2004, at 8:56 PM, Lever, Charles wrote: > charles- > >>> [ snip ] >>> >>> >> We'll see if this disappears after we've finished migrating >> the network >> to 1Gbps >> >>> In these environments you *must* use TCP, since that has congestion >>> control capabilities baked into the protocol... > > what matters here is flow control. UDP simply doesn't have it. > > for gigabit Ethernet, you need flow control at the link level and > at the transport level. so once you have gigabit infrastructure, > be sure you have enabled full gigabit flow control on your servers > and on your switches. then you should use TCP and not UDP so you > have transport layer flow control. > Charles and Trond, thanks for the advices ! I've upgraded the network cards of the linux machines to Gigabit ethernet cards and used tcp for my nfs mounts. The problem disappeard so far so i'm pretty happy ! However i'm very disappointed by the throuhput i'm getting on the Gbps network : When i scp a 473991492 bytes file from one machine to the other, i'm getting 14Mbps throughput, which is far less that 1Gpbs theoretical. Looking at the network card info ( i'm using the sk98lin driver ) i've : Mar 18 14:49:15 omnirot kernel: eth0: network connection up using port A Mar 18 14:49:15 omnirot kernel: speed: 1000 Mar 18 14:49:15 omnirot kernel: autonegotiation: yes Mar 18 14:49:15 omnirot kernel: duplex mode: full Mar 18 14:49:15 omnirot kernel: flowctrl: symmetric Mar 18 14:49:15 omnirot kernel: role: slave Mar 18 14:49:15 omnirot kernel: irq moderation: disabled Mar 18 14:49:15 omnirot kernel: scatter-gather: enabled The switch i'm using is a Netgear GS516T i've tried transferring to/from 2 linux machines connected to this switch using the exact same kernel and network cards and also to Mac Xserv using a Gbps card . They both yield the same low throuput. Any hint of what could be causing this ? Thanks in advance. Charles-Edouard Ruault Idtect SA tel: +33-1-42-81-81-84 fax: +33-1-42-81-82-21 http://www.idtect.com ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ NFS maillist - NFS@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/nfs