From mboxrd@z Thu Jan 1 00:00:00 1970 From: Wido den Hollander Subject: Re: some questions about rbd Date: Thu, 25 Oct 2012 09:27:02 +0200 Message-ID: <5088E9C6.9080403@widodh.nl> References: <50861B39.8010008@inktank.com> <5086386D.2030002@inktank.com> <5086D830.5010505@inktank.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from smtp02.mail.pcextreme.nl ([109.72.87.138]:45474 "EHLO smtp02.mail.pcextreme.nl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757919Ab2JYH1H (ORCPT ); Thu, 25 Oct 2012 03:27:07 -0400 In-Reply-To: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: jie sun <0maidou0@gmail.com> Cc: Josh Durgin , ceph-devel On 10/25/2012 08:25 AM, jie sun wrote: > I'm are about to support a service that vm cloud mount new device dynamically. > I'm afraid that sometimes the vm crashed(not totally crashed but just > can't be connected with, or the host can't be connected so that I > can't remove the vm at all) but the block device is still in use, and > the daemon that write or read block device is still alive. > Maybe in the situation that we meet a spiteful client and do some > operations not allowed. > > Is it necessary to consider these situations? The wather of the image > is on the mon or the osd? Can you tell me what is the priciple of the > watcher? some tcp connection(the vm to the osd)? > If I were you I'd switch to using the build in RBD functionalities of Qemu instead of mounting a block device and running your virtual machine from that device. That would also be a lot easier since you don't have to map/unmap a device. You can also kill a VM much easier since it is not that much tied into the kernel. You should take a look at libvirt and Qemu. Wido > Thank you! > > > 2012/10/24 Josh Durgin : >> Currently there's no way to force deletion of a RADOS object that >> has watchers. Clients with an rbd image open watch the header >> object, hence the error you saw. >> >> There's a feature request open for this at the RADOS level: >> http://www.tracker.newdream.net/issues/2339 >> >> I'm curious about why you want to delete an image even >> if a vm is still using it. What's the context in which this >> becomes necessary? >> >> Josh >> >> >> On 10/22/2012 11:25 PM, Dan Mick wrote: >>> >>> I don't think we support that; I'm not sure how hard it would be. >>> Somehow we'd have to keep knowledge of the open image handle around to >>> fail further I/O requests, it seems. Josh? >>> >>> On 10/22/2012 10:44 PM, jie sun wrote: >>>> >>>> Yes. I want to delete the image by force, even the vm is still using it. >>>> >>>> 2012/10/23 Dan Mick : >>>>> >>>>> >>>>> >>>>> On 10/22/2012 08:10 PM, jie sun wrote: >>>>>> >>>>>> >>>>>> Hi, >>>>>> >>>>>> I create a image and map it to a virtual machine,and then mkfs and >>>>>> mount >>>>>> it. >>>>>> >>>>>> I want to remove the image from mon server forcely, but it says >>>>>> "Removing image: 99% complete...failed. >>>>>> delete error: image still has watchers >>>>>> This means the image is still open or the client using it crashed. >>>>>> Try again after closing/unmapping it or waiting 30s for the crashed >>>>>> client to timeout." >>>>>> >>>>>> I want to know is there any other way to remove it by force? >>>>> >>>>> >>>>> >>>>> Are you saying the vm still has the device open and is trying to use it? >>>>> But even so, you want the image to be deleted? >>>>> >>> -- >>> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in >>> the body of a message to majordomo@vger.kernel.org >>> More majordomo info at http://vger.kernel.org/majordomo-info.html >> >> > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >