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

Re: [Xen-users] propagating blktap partition changes


  • To: xen-users@xxxxxxxxxxxxxxxxxxx
  • From: Maarten Vanraes <maarten@xxxxx>
  • Date: Thu, 5 Feb 2009 15:35:18 +0100
  • Cc:
  • Delivery-date: Thu, 05 Feb 2009 06:40:06 -0800
  • List-id: Xen user discussion <xen-users.lists.xensource.com>

Op Thursday 05 February 2009 15:27:55 schreef Pasi Kärkkäinen:
> On Thu, Feb 05, 2009 at 01:59:57PM +0100, Maarten Vanraes wrote:
> > Op Thursday 05 February 2009 13:05:01 schreef Pasi Kärkkäinen:
> > > On Thu, Feb 05, 2009 at 11:52:56AM +0100, Maarten Vanraes wrote:
> > > > Hi,
> > > >
> > > > I have 2 questions:
> > > >
> > > >  1. I was not able to let the guest know that the partition size was
> > > > changed. 2. I was not able to use block-attach with the same
> > > > backendfile to different machines (eg: guest + dom0)
> > > >
> > > > I've read that block-reconfigure could've been used for this, except
> > > > that it wasn't designed to work for filebased vbd on PVs, only for
> > > > switching CDROMs in HVM
> > > >
> > > > When you use a local disk on a machine, you can use fdisk to change
> > > > partitions and sizes, it then rereads the partition table to reflect
> > > > those changes.
> > > >
> > > > Since i'm not using a disk, but directly partitions, isn't it
> > > > possible to just give a signal to the guest that it should recheck
> > > > the partition? can be from the guest itself.
> > > >
> > > > Some kind of workaround?
> > >
> > > Sorry, this is not possible.
> > >
> > > There's a a feature request about it for Xen 3.4.
> > >
> > > Dunno if it will be implemented.
> > >
> > > -- Pasi
> >
> > Yes, i read your thread, i was hoping for some kind of workaround, some
> > kind of "thing" i can execute on the guest, that would make the kernel
> > recheck it's partition size or something.
>
> The problem is not about rereading the partition table; this can be done
> using for example "blockdev --rereadpt".
>
> The problem is to get the disk device (/dev/xvd*) resized.. ie. getting the
> kernel to see the DISK has been resized.
>
> After the disk has been resized, partition table can be resized after that.
>
> Disk resizing on the fly is not supported today with Xen.

Well I understand that it is inside the kernel, but, i'm not using disks, i'm 
using partitions directly:

e.g.:

'tap:aio:/opt/data/file.img, sda1, w'

I can make the partition file larger on dom0 no problem, i could even resize 
the partition live on dom0 no problem.

But (i guess the kernel) doesn't see the partition as larger, even with forced 
operations, i still get read errors at the end.

I'm not even sure the blktap device detected this change in size on dom0.

But if i'm using directly partitions, maybe there is some code that i can 
execute that does kernel calls or something to refresh it's device. ie: a 
workaround.

I mean, hotplug seems to be able to do it just fine. ie: memory is detected as 
larger, i imagine there is such code for block devices as well, no? So i 
figured there should be something i can do in the guest (maybe even on kernel 
level) that could force the re-detection of the device or something.

Of course if blktap can't even pass on the information to the guest, this 
would not work at all...

Sincerely,

Maarten

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.