qemu-block
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-block] 回复: [qemu-block]Can qemu close the image files with gu


From: Stefan Hajnoczi
Subject: Re: [Qemu-block] 回复: [qemu-block]Can qemu close the image files with guestOS running?
Date: Mon, 27 Feb 2017 10:55:21 +0000
User-agent: Mutt/1.7.1 (2016-10-04)

On Thu, Feb 23, 2017 at 06:21:00AM +0800, 丁强 wrote:
> 在2017年02月22日 23:39,Stefan Hajnoczi 写道:
> On Wed, Feb 22, 2017 at 10:54:50AM +0800, 丁强 wrote:
> > > I deploy the VMs on filesystem which based on the iscsi LUNs. When the 
> > > network of san is corrupted occasionally, the filesystem will be 
> > > remounted read-only by OS with default policy. If you want to remount the 
> > > filesystem to readwrite, you must shutdown or migrate firstly. Can we 
> > > recover from this sense without VMs operation, just like vmware without 
> > > VMCP configuration.
> >
> > "mount -o remount,rw /dir" on the host does not work?
> 
> When there are some files opened, it doesn't work with busy errors.

You are right.  I thought it would work but am getting EBUSY here too.

If the file system has become read-only it poses a serious problem
because QEMU may be unable to quiesce the disk (bdrv_drain_all()).
Workarounds like live migration or savevm may not work due to this.

There is currently no simple way to recover from this besides shutting
down QEMU, remounting the file system, and starting QEMU again.

It would be great to have a proper solution to this problem.  It's
important to keep in mind that continuing to run the VM after a disk
failure carries a risk: the resumed disk may not be in the state that
the VM expects.  This can cause I/O errors or weird behavior inside the
guest.  So I think that the guest doing I/O during the failure must
experience a disk error so that it has a chance to recover, remount,
fsck, etc.

Stefan

Attachment: signature.asc
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]