Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266441AbUFUUDs (ORCPT ); Mon, 21 Jun 2004 16:03:48 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266442AbUFUUDs (ORCPT ); Mon, 21 Jun 2004 16:03:48 -0400 Received: from parcelfarce.linux.theplanet.co.uk ([195.92.249.252]:64473 "EHLO www.linux.org.uk") by vger.kernel.org with ESMTP id S266441AbUFUUDp (ORCPT ); Mon, 21 Jun 2004 16:03:45 -0400 Date: Mon, 21 Jun 2004 16:56:50 -0300 From: Marcelo Tosatti To: "Ramy M. Hassan" Cc: linux-kernel@vger.kernel.org Subject: Re: kswapd problem Message-ID: <20040621195650.GA13944@logos.cnet> References: <20040612153247.13279.qmail@gawab.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20040612153247.13279.qmail@gawab.com> User-Agent: Mutt/1.5.5.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1564 Lines: 38 On Sat, Jun 12, 2004 at 03:32:47PM +0000, Ramy M. Hassan wrote: > > kswapd and kupdated are causing our production server to > completely freezeup for few seconds every now and then. > The server is running kernel 2.4.26SMP on a Dual Xeon 2.20GHz > with 4GB RAM, 900GB FC RAID Qlogic HBA using driver qla2300.o > and reiserfs. > The RAID filesystem contains millions of files in thousands of > directories. > The system is under fair load. Normally the load avarage is > about 3 and everything works properly, but suddenly the system > stops responding except to ping, and stay freezed for about 20 > seconds, during that time I can not even type anything, then the > system becomes responsive again and I see the load avarge over > 250 and starts to decrease till it is back to 3 , then few > minutes later that same thing is repeated. > I noticed that at the time of the freezups both kswapd and > kupdated are the most active processes each consuming over 30% > of the CPU ( kswapd is usually more than kupdated ) Hi Ramy, Can you get us some more data when this happens? What are the size's of the page lists (/proc/meminfo) when the freeze happens, can you capture that? Also leave vmstat running in the background. If you are willing to debug I'm sure we will be able to find the reason for the problem. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/