From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Tue, 27 Sep 2016 14:56:45 +0100 From: "Bryn M. Reeves" Message-ID: <20160927135644.GA29650@localhost.localdomain> MIME-Version: 1.0 Content-Disposition: inline Subject: [linux-lvm] dmioscope: visualise IO distribution for DM devices Reply-To: LVM general discussion and development List-Id: LVM general discussion and development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , List-Id: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: dm-devel@redhat.com Cc: linux-lvm@redhat.com I wanted to give a quick plug for a little tool I've been working on for the last few weeks: dmioscope is a python script that uses device-mapper statistics to generate histograms of IO distribution on a DM device: # python dmioscope.py -C READS -b 6 3 0 vg_hex/root Initialised READS adaptive histogram with 6 bins, min_size=1677721. vg_hex/root: current IO distribution 0.0 2000.0 4000.0 6000.0 8000.0 +------------------------------------------------------------- 0.0B |####### 5.3GiB |##################################################### 10.7GiB |##### 16.0GiB |### 21.3GiB | 26.7GiB | 90.00% of IO reaches 50.00% of disk. [...] (watch reads, six bins, interval of 3s, carry on forever) It's useful for finding and tracking hot-spots in real time, and for getting a general picture of where IO is going to on a device. I'm hoping it can also be used to get an idea of the benefits that can be achieved with caching. I've pushed it to GitHub - if anyone finds any bugs, or has any suggestions, I'm very interested to hear them: there's a README.md with usage information and more details on the histogram implementation on the main page: https://github.com/bmr-cymru/dmioscope Regards, Bryn.