From mboxrd@z Thu Jan 1 00:00:00 1970 From: Yuehai Xu Subject: Does tapdisk2 dispatch requests from different DomUs? Date: Wed, 26 Jan 2011 10:36:55 -0500 Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Return-path: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: xen-devel@lists.xensource.com Cc: George Dunlap , Jeremy Fitzhardinge , keir@xen.org, yhxu@wayne.edu List-Id: xen-devel@lists.xenproject.org Hi All, Suppose there are two DomUs, and each one has a corresponding tapdisk2 process to dispatch requests. Theoretically, different tapdisk should only dispatch requests from its corresponding DomU. However, I notice that a certain tapdisk would dispatch requests from other DomUs, is it normal? Below is the background: Dom1.cfg: kernel="/boot/vmlinuz-2.6.31.13-xen-guest_32bit_ubuntu" name = "dom1" vif = ['mac=00:1A:2A:4A:6A:8A, bridge=xenbr0, ip=192.168.1.111'] disk = ['tap2:aio:/guest_os/dom1/img,xvda1,w', 'tap2:aio:/guest_os/dom1/swap,xvda2,w', 'tap2:aio:/mnt/sdb1/xen/img,xvdc,w'] root = "/dev/xvda1" extra = "fastboot enforcing=0 console=hvc0" Dom2.cfg: kernel="/boot/vmlinuz-2.6.31.13-xen-guest_32bit_ubuntu" name = "dom2" vif = ['mac=00:1B:2B:4B:6B:8B, bridge=xenbr0, ip=192.168.1.112'] disk = ['tap2:aio:/guest_os/dom2/img,xvda1,w', 'tap2:aio:/guest_os/dom2/swap,xvda2,w', 'tap2:aio:/mnt/sdb7/xen/img,xvdb,w'] root = "/dev/xvda1" extra = "fastboot enforcing=0 console=hvc0" ps -ef | grep tapdisk | grep sdb : root 7225 1 0 09:51 ? 00:00:18 /usr/sbin/tapdisk2 -n aio:/mnt/sdb7/xen/img root 8670 1 0 10:13 ? 00:00:00 /usr/sbin/tapdisk2 -n aio:/mnt/sdb1/xen/img When I run postmark which creates small random writes in both Doms, each tapdisk dispatches requests both from Dom1 and Dom2. In this way, it is impossible to do QoS for disk I/O since it will never be known which requests are from which DomUs. Is this a known problem or is there something wrong that I have done? I appreciate your help. Thanks, Yuehai