From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from merlin.infradead.org ([205.233.59.134]:37634 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752076Ab3EVI54 (ORCPT ); Wed, 22 May 2013 04:57:56 -0400 Date: Wed, 22 May 2013 10:57:36 +0200 From: Jens Axboe Subject: Re: output - aggregate read bandwitdh Message-ID: <20130522085736.GE9714@kernel.dk> References: <20130518065341.GJ19781@kernel.dk> <20130518070110.GK19781@kernel.dk> <20130518122211.GM19781@kernel.dk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: fio-owner@vger.kernel.org List-Id: fio@vger.kernel.org To: Edoardo Comar Cc: fio@vger.kernel.org On Wed, May 22 2013, Edoardo Comar wrote: > > > > I just tried the code from git on my personal macbook. > > I replaced libaio with posixaio in the config file and > > the ultra-high aggrb read is there too (files being created). > > > > is caused by the contribution to the aggregate stats > > of the second of the two query threads [queryB] > > > > > > Run status group 0 (all jobs): > > READ: io=16350MB, aggrb=277927KB/s, minb=10KB/s, maxb=278418KB/s, > > mint=60001msec, maxt=60240msec > > WRITE: io=10164KB, aggrb=168KB/s, minb=8KB/s, maxb=160KB/s, > > mint=60088msec, maxt=60240msec > > Hi, are there any insight on the READ aggrb issue? > I'm fairly confident it's a bug now that I see it happening even on > different platforms (mac and linux) It would help a lot if you could do a git bisect and reliably find out which commit caused the problem. The procedure is something like this, assuming you are in the fio git repository: $ git bisect start $ git bisect good fio-2.0.13 $ git bisect bad master Then make clean and make fio, run the job file: $ make clean; make $ ./fio Run the job file as many times as is required to reliably tell whether this is a good or a bad build. Lets assume the first one is good, you would then do: $ git bisect good and repeat the make clean && make && run of job file. If the result was bad, you'd do git bisect bad. You continue this process until it tells you which one is the offending commit. Then you mail that here :-) -- Jens Axboe