From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ryan Harper Subject: Re: [PATCH 0/6] xen, xend, tools: Add NUMA support to Xen Issues on the ES7000 Date: Thu, 17 Aug 2006 11:47:43 -0500 Message-ID: <20060817164743.GD1694@us.ibm.com> References: <94C8C9E8B25F564F95185BDA64AB05F6040C8D2E@USTR-EXCH5.na.uis.unisys.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <94C8C9E8B25F564F95185BDA64AB05F6040C8D2E@USTR-EXCH5.na.uis.unisys.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: "Subrahmanian, Raj" Cc: Ian Pratt , Ryan Harper , xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org * Subrahmanian, Raj [2006-08-16 22:29]: > Ryan, > I put in the NUMA patches 1-6 on changeset 10946 (the patches wouldn't > apply cleanly on the current tip). I ran with both 1 cell and 2 cells. I'll be sending an update soon. > I have attached the complete output. > I got the heap and the numa info before and after running xm top (which > slowed the system down dramatically as mentioned previously). > http://lists.xensource.com/archives/html/xen-devel/2006-08/msg00348.html > As mentioned previously, the system does not die (I can get the debug > info, I can ping etc). But the system slows down dramatically, so > ssh-ing into the box is impossible. > Let me know what else I can try out/look at etc. I can't reproduce the hang, however, there are some issues that need fixing up with libxenstat and the physinfo hypercall. I'll include those fixes in when I refresh to tip today. If you can, I'd be interested in seeing what process in dom0 is chewing up the cpu if you can get that info. I doubt this is an issue, but can you repro with sedf scheduler as well? And while I don't know much about the credit scheduler, dumping the runqueues 'r' keyhandler might show something to the folks who know the credit scheduler better. Oh, and if you have a log of xen/dom0 booting on your 1 and 2 cell box with out the numa patches on the same changeset, I'd like to see that to compare output. -- Ryan Harper Software Engineer; Linux Technology Center IBM Corp., Austin, Tx (512) 838-9253 T/L: 678-9253 ryanh@us.ibm.com