From: Shehjar Tikoo Subject: Re: Errors during a simple mount storm test Date: Fri, 05 Feb 2010 02:12:51 +0530 Message-ID: <4B6B314B.4050408@gluster.com> References: <4B6B2AE5.1070007@gluster.com> <1265315573.3317.50.camel@localhost> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Cc: linux-nfs@vger.kernel.org To: Trond Myklebust Return-path: Received: from zp3.zcs.datasyncintra.net ([208.88.241.29]:34959 "EHLO zp3.zcs.datasyncintra.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754673Ab0BDUmu (ORCPT ); Thu, 4 Feb 2010 15:42:50 -0500 In-Reply-To: <1265315573.3317.50.camel@localhost> Sender: linux-nfs-owner@vger.kernel.org List-ID: Trond Myklebust wrote: > On Fri, 2010-02-05 at 01:45 +0530, Shehjar Tikoo wrote: >> Hi All >> >> The shell command below is probably not applicable to >> all real world setups but I wanted to check it out nonetheless. >> The problem is with the following command that is supposed >> run a mount-umount loop: >> >> $ while [ true ]; do echo sending; mount localhost:/test -o >> noacl,nolock,tcp,mountproto=tcp /mnt;umount /mnt; done >> >> For the first hundred iterations, it works fine but then starts >> spewing errors with the mount command failing. For eg. >> the output looks like below: >> >> sending >> sending >> sending >> sending >> sending >> sending >> sending >> sending >> sending >> sending >> sending >> sending >> mount.nfs: mount system call failed >> umount: /mnt: not mounted >> sending >> mount.nfs: mount system call failed >> umount: /mnt: not mounted >> sending >> mount.nfs: mount system call failed >> umount: /mnt: not mounted >> sending >> mount.nfs: mount system call failed >> umount: /mnt: not mounted >> >> Any ideas why the mount sequence stops abruptly? >> There dont seem to be any relevant messages in the >> dmesg output either. >> >> The client and server are on the same machine: >> Linux indus 2.6.31-17-generic Ubuntu SMP x86_64 GNU/Linux >> >> I have verified through wireshark that the server is not >> receiving any mount requests once the mount command start >> failing. Instead there are only successful GETPORT requests >> for the MOUNTv3 program. > > "netstat -t" will probably show you that all your free ports in the > range 0-1023 are now in the TIME_WAIT state. This again means that you > have to wait a couple of minutes for the TIME_WAITs to expire before you > can try again. > > So basically, it is a case of your test being flawed. Thanks! I'll to resort to using noresvport in that case. -Shehjar > > Trond >