From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Madovsky" Subject: Re: [ANNOUNCE] 3.0-rc7-rt0 Date: Wed, 20 Jul 2011 11:59:24 -0400 Message-ID: <7735DD508B2B433B9B13F223209C5F85@e1705> References: <4E26B721.9020103@steinhoff.de> Mime-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-8859-1"; reply-type=original Content-Transfer-Encoding: 7bit To: "linux-rt-users" Return-path: Received: from ip-67-205-80-138.static.privatedns.com ([67.205.80.138]:32168 "EHLO node250.e-blokos.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751584Ab1GTP7f (ORCPT ); Wed, 20 Jul 2011 11:59:35 -0400 Received: from e1705 (localhost.localdomain [127.0.0.1]) by node250.e-blokos.com (8.14.3/8.14.3) with SMTP id p6KFxYSF006347 for ; Wed, 20 Jul 2011 11:59:35 -0400 Sender: linux-rt-users-owner@vger.kernel.org List-ID: Hi Thomas, ----- Original Message ----- From: "Thomas Gleixner" To: "Madovsky" Cc: "linux-rt-users" Sent: Wednesday, July 20, 2011 11:56 AM Subject: Re: [ANNOUNCE] 3.0-rc7-rt0 > Franck, > > On Wed, 20 Jul 2011, Madovsky wrote: > >> Hi Thomas, >> >> I just compiled the kernel with the patch on my whole cluster >> of 6 fedora10 and the results are very stable and smooth. > > Thanks for the report ! you're welcome ! :D >> the only little thing is the cluster software heartbeat / pacemaker >> takes 20% of cpu that 11% before on a dual xeon nocona with >> linux command "top". > > That might be related to the new mechanism how we handle softirqs. If > the heartbeat/pacemaker app is using networking a lot then this might > explain it. > > Thanks, > > tglx yes it uses constantly the network (1GB hub between the nodes) for high availability. Thanks for the explaination Franck