From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Christie Subject: Re: [PATCH 0/5] target/rbd: distributed lun reset support Date: Tue, 05 May 2015 23:56:22 -0500 Message-ID: <55499EF6.3080906@redhat.com> References: <1430819778-24483-1-git-send-email-mchristi@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Return-path: Received: from mx1.redhat.com ([209.132.183.28]:51257 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751252AbbEFE4X (ORCPT ); Wed, 6 May 2015 00:56:23 -0400 In-Reply-To: <1430819778-24483-1-git-send-email-mchristi@redhat.com> Sender: ceph-devel-owner@vger.kernel.org List-ID: To: ceph-devel@vger.kernel.org, target-devel@vger.kernel.org On 05/05/2015 04:56 AM, mchristi@redhat.com wrote: > The following patches allow drivers like libceph/rbd to distribute > execution of scsi task management functions across multiple nodes > (still working on issues with distributed PRs). > > In this patchset, I added a new type of callout/module for these > cluster aware modules. The idea is that something like drbd could maybe do > something optimized for them, or nbd could, or you could use dlm > plus some file system, or do something cluster FS specific. For rbd > we are making rados/libceph calls. > > I am mostly looking for feedback on the hooks/callouts approach. I was > not sure if we wanted to maybe instead do somethign like add new callouts > to the request_queue (request_queue->reset_q) or block_device or > add some new block layer REQ_TYPEs (REQ_TYPE_DEVICE_RESET or > REQ_TYPE_PGR_REGISTER, etc). In this patchset I just added new > callouts that modules can implement, because I thought this was most > similar to how we do fabric modules and so it was a ok fit. > Oh shoot. Sorry. Forgot to cc the lio list on the first set of ceph/rbd patches which this patchset is built on. They are here: http://thread.gmane.org/gmane.comp.file-systems.ceph.devel/24553