From mboxrd@z Thu Jan 1 00:00:00 1970 From: Kent Overstreet Subject: Tiering in bcachefs Date: Mon, 12 Sep 2016 18:17:11 -0800 Message-ID: <20160913021711.oxowdbp2sim5xwaq@kmo-pixel> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mail-pf0-f177.google.com ([209.85.192.177]:34230 "EHLO mail-pf0-f177.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752199AbcIMCRP (ORCPT ); Mon, 12 Sep 2016 22:17:15 -0400 Received: by mail-pf0-f177.google.com with SMTP id p64so58473788pfb.1 for ; Mon, 12 Sep 2016 19:17:15 -0700 (PDT) Received: from kmo-pixel (213-168-42-72.gci.net. [72.42.168.213]) by smtp.gmail.com with ESMTPSA id p74sm27199464pfk.68.2016.09.12.19.17.12 for (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Mon, 12 Sep 2016 19:17:13 -0700 (PDT) Content-Disposition: inline Sender: linux-bcache-owner@vger.kernel.org List-Id: linux-bcache@vger.kernel.org To: linux-bcache@vger.kernel.org Just wrote a long post explaining all about tiering in bcachefs: https://www.patreon.com/posts/all-about-6751413 Also, tiering should now be working! I'd appreciate help testing it. Notes: Chris Halse Rogers (RAOF in the #bcache IRC channel) has been testing it. He has been seeing an intermittent deadlock while copying large amounts of data, which may or may not be tiering related: if anyone else hits it, I'd really appreciate if you could grab backtraces. Do a "echo t > /proc/sysrq-trigger" and then grab the full dmesg log - that should be enough to figure out what the deadlock is. I haven't gotten to device add/remove yet, don't try it - you'll need to create a new filesystem: bcache format --tier 0 --tier 1 Also, if you've already been playing with tiering, I highly recommend you create a new filesystem. One of the bugs was an issue with which pointers are considered dirty - in an extent with a tier 0 pointer and a tier 1 pointer, the tier 0 pointer was being considered dirty instead of the tier 1 pointer (with predictably bad consequences; tier 0 would fill up because the tiering thread wasn't actually making anything on tier 0 clean). On a fixed kernel, as extents are migrated it'll end up fixing which pointers are considered dirty which may or may not trigger assertions as which pointers are considered dirty changes.