From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261337AbTHSTsE (ORCPT ); Tue, 19 Aug 2003 15:48:04 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261274AbTHSTrN (ORCPT ); Tue, 19 Aug 2003 15:47:13 -0400 Received: from web41411.mail.yahoo.com ([66.218.93.77]:58765 "HELO web41411.mail.yahoo.com") by vger.kernel.org with SMTP id S261337AbTHSTqp (ORCPT ); Tue, 19 Aug 2003 15:46:45 -0400 Message-ID: <20030819194644.26776.qmail@web41411.mail.yahoo.com> Date: Tue, 19 Aug 2003 12:46:44 -0700 (PDT) From: Rock Gordon Subject: init consumes 99% cpu, syslog Z To: linux-kernel@vger.kernel.org MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hi, I am using linux-2.4.18 (vanilla, from ftp.kernel.org) on a RedHat 7.3 release. The filesystem is XFS. When doing heavy I/O (16 processes each writing or reading between 2 and 5 GB of data to a 2TB XFS filesystem), I see wierd unsolvable problems - Firstly, init all of a sudden starts consuming upwards of 99% cpu (the profiler shows that it spends most of it's time in the functions send_sig_info()/force_sig_info()). Pretty soon (10-15 seconds or so), syslogd becomes a zombie, with init still spinning in R mode. Absolutely nothing in /var/log/messages; and dmesg shows nothing either. No visible barfing. And this goes on and on. Any process I run after this keeps getting segfaults, and also ends up becoming a zombie. This problem repeats exactly on other identical machines. The system is a dual-CPU 2.4Ghz Dell 2650 machine and the problems show up without hyperthreading too. The problem shows up without XFS too. Any ideas/suggestions? Thanks and Regards, Rocky PS: I am not on the list. __________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com