From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1762734AbYFGBao (ORCPT ); Fri, 6 Jun 2008 21:30:44 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1762080AbYFGBNG (ORCPT ); Fri, 6 Jun 2008 21:13:06 -0400 Received: from one.firstfloor.org ([213.235.205.2]:32892 "EHLO one.firstfloor.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1761824AbYFGBNE (ORCPT ); Fri, 6 Jun 2008 21:13:04 -0400 Message-ID: <4849E09D.6090804@firstfloor.org> Date: Sat, 07 Jun 2008 03:13:01 +0200 From: Andi Kleen User-Agent: Thunderbird 1.5.0.12 (X11/20060911) MIME-Version: 1.0 To: Scott Wiersdorf CC: Balbir Singh , linux-kernel@vger.kernel.org, akpm@linux-foundation.org, matt@bluehost.com Subject: Re: [PATCH 2.6.25-4] getdelays.c: signal handling for log rotation References: <20080605194334.GA56830@perlcode.org> <87d4muu0k2.fsf@basil.nowhere.org> <20080606204704.GA4330@perlcode.org> In-Reply-To: <20080606204704.GA4330@perlcode.org> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Scott Wiersdorf wrote: > On Fri, Jun 06, 2008 at 03:59:09PM +0200, Andi Kleen wrote: >> Scott Wiersdorf writes: >> >>> This adds a USR1 signal handler to getdelays.c, which causes getdelays >>> to close its logfile and reopen it (if '-w logfile' is >>> specified). This is useful in situations when getdelays is running for >>> a long time (i.e, the log file growing) and you need to rotate the >>> logs but don't want to lose any log data. >> You could do the same by sending SIGSTOP; copy file; truncate file; SIGCONT > > Actually, I was wrong in my previous reply. Sorry for my error. The > above will work fine (no data loss; I guess the data queues somewhere > in some magic way?) netlink has a socket buffer like all other buffer. I think it's around 128K. > I still think a single handler is elegant enough though, and works > better with many log rotation systems that want to send a single > signal to a pid (it's what we need where I'm working now, hence the > patch). Well in general it would be cool if there was nicer free userland for the obscure but useful delay accounting. The code in Documentation is really not much more than a example. Do you have something downloadable? -Andi