Re: Real Device with BHyve

2014-01-03 Thread Andrea Brancatelli
Forgive me, but I can't understand what you mean.

Are we talking of using something like /dev/cciss (just to say...) instead
of /dev/da2 as the device shared with the VM?

Won't the VM and the real system clash in using the same device?



On Thu, Jan 2, 2014 at 9:29 PM, Nikolai Lifanov lifa...@mail.lifanov.comwrote:

 On 01/02/14 15:22, freebsd-virtualization-requ...@freebsd.org wrote:
  Hello everybody.
 
  I'm doing some experiments with bhyve on 10.0-RC3 and I got stuck at a
  certain point.
 
  I was trying to have a VM use a direct device (/dev/da2) instead of a
 disk
  image. I was trying it in order to understand if there was any real
  performance difference between using a raw drive or an image-disk on the
  same drive.
 
  Well, the machine starts ok but when the child FreeBSD starts
  installation something strange happens. When I get to the partitioning
  screen I can see the device avaiable as /dev/vtdb0 with the correct size
  and such. I choose autopartitioning, the installer writes the partition
  table but when it start to write /dev/vtdb0p2 a very cryptic error
 appears
  about being unable to write - sorry, did not write it down.
 
  The installer then stops.
 
  If I do a fdisk /dev/vtdb0 in the VM I can see the GPT partition being
  there. If I do a fdisk /dev/da2 on the host machine, I can see the GPT
  partition as well, but the VM just doesn't want to write on it.
 
  I even tried changing kern.geom.debugflags=16 as I thought the host
 machine
  could be locking somehow the drive, but that didn't seem to make any
  difference. I know it was a lame check but I was out of ideas.
 
  So I just wanted to understand if such a scenario is supposed to be
  supported
 
  What I was thinking of, for example, was of having an external iSCSI
 device
  connected on the hostmachine mapped as a virtual disk for a specific VM,
 in
  order to speed the VM disk performances.
 
 
  Just another quick question... I have seen some improvements by having
 the
  VM's virtual disk on ZFS against UFS. Is it just me or is there any real
  improvement by using ZFS?
 
  Thanks a lot.
 
 
  -- *Andrea BrancatelliSchema 31 S.r.l. - Socio UnicoResponsabile ITROMA
  - FIRENZE - PALERMO ITALYTel: +39. 06.98.358.472* *Cell: +39
  331.2488468Fax: +39. 055.71.880.466Societ? del Gruppo SC31 ITALIA*

 I'm not answering your question precisely, but can you pass through the
 disk controller to the virtual machine instead? I also know that zvol
 and iscsi backends work, at least the last time I checked.

 - Nikolai Lifanov
 ___
 freebsd-virtualization@freebsd.org mailing list
 http://lists.freebsd.org/mailman/listinfo/freebsd-virtualization
 To unsubscribe, send any mail to 
 freebsd-virtualization-unsubscr...@freebsd.org




-- 




*Andrea BrancatelliSchema 31 S.r.l. - Socio UnicoResponsabile ITROMA -
FIRENZE - PALERMO ITALYTel: +39. 06.98.358.472*

*Cell: +39 331.2488468Fax: +39. 055.71.880.466Società del Gruppo SC31
ITALIA*
___
freebsd-virtualization@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-virtualization
To unsubscribe, send any mail to 
freebsd-virtualization-unsubscr...@freebsd.org

Re: Real Device with BHyve

2014-01-03 Thread Peter Grehan

Hi Andrea,

Well, the machine starts ok but when the child FreeBSD starts
installation something strange happens. When I get to the partitioning
screen I can see the device avaiable as /dev/vtdb0 with the correct size
and such. I choose autopartitioning, the installer writes the partition
table but when it start to write /dev/vtdb0p2 a very cryptic error appears
about being unable to write - sorry, did not write it down.

The installer then stops.

If I do a fdisk /dev/vtdb0 in the VM I can see the GPT partition being
there. If I do a fdisk /dev/da2 on the host machine, I can see the GPT
partition as well, but the VM just doesn't want to write on it.

I even tried changing kern.geom.debugflags=16 as I thought the host machine
could be locking somehow the drive, but that didn't seem to make any
difference. I know it was a lame check but I was out of ideas.

So I just wanted to understand if such a scenario is supposed to be
supported


 It is. Been a while since I've done this but will try a repro. Other 
folk have supported success using zvols so I'd assumed it was working.



What I was thinking of, for example, was of having an external iSCSI device
connected on the hostmachine mapped as a virtual disk for a specific VM, in
order to speed the VM disk performances.


 Yes, that's one of the scenarios in mind.


Just another quick question... I have seen some improvements by having the
VM's virtual disk on ZFS against UFS. Is it just me or is there any real
improvement by using ZFS?


 Difficult question to answer - probably workload-dependent.

later,

Peter.
___
freebsd-virtualization@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-virtualization
To unsubscribe, send any mail to 
freebsd-virtualization-unsubscr...@freebsd.org


Re: Real Device with BHyve

2014-01-03 Thread Michael Gmelin


On Fri, 03 Jan 2014 09:43:24 -0800
Peter Grehan gre...@freebsd.org wrote:

 Hi Andrea,
  Well, the machine starts ok but when the child FreeBSD starts
  installation something strange happens. When I get to the
  partitioning screen I can see the device avaiable as /dev/vtdb0
  with the correct size and such. I choose autopartitioning, the
  installer writes the partition table but when it start to
  write /dev/vtdb0p2 a very cryptic error appears about being unable
  to write - sorry, did not write it down.
 
  The installer then stops.
 
  If I do a fdisk /dev/vtdb0 in the VM I can see the GPT partition
  being there. If I do a fdisk /dev/da2 on the host machine, I can
  see the GPT partition as well, but the VM just doesn't want to
  write on it.
 
  I even tried changing kern.geom.debugflags=16 as I thought the host
  machine could be locking somehow the drive, but that didn't seem to
  make any difference. I know it was a lame check but I was out of
  ideas.
 
  So I just wanted to understand if such a scenario is supposed to be
  supported
 
   It is. Been a while since I've done this but will try a repro.
 Other folk have supported success using zvols so I'd assumed it was
 working.

If it helps, I had the same problem using zvol as well (booting the
installer and install on the raw device) on 10-CURRENT from mid 2013.

I didn't try with a more recent release. I ended up installing FreeBSD
from the outside (create partitions and untar base.txz, kernel.txz etc.
- afaik the same thing the Dexter's vm0 script does now) and could then
boot into it just fine.

-- 
Michael Gmelin
___
freebsd-virtualization@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-virtualization
To unsubscribe, send any mail to 
freebsd-virtualization-unsubscr...@freebsd.org


Re: Real Device with BHyve

2014-01-02 Thread Nikolai Lifanov
On 01/02/14 15:22, freebsd-virtualization-requ...@freebsd.org wrote:
 Hello everybody.
 
 I'm doing some experiments with bhyve on 10.0-RC3 and I got stuck at a
 certain point.
 
 I was trying to have a VM use a direct device (/dev/da2) instead of a disk
 image. I was trying it in order to understand if there was any real
 performance difference between using a raw drive or an image-disk on the
 same drive.
 
 Well, the machine starts ok but when the child FreeBSD starts
 installation something strange happens. When I get to the partitioning
 screen I can see the device avaiable as /dev/vtdb0 with the correct size
 and such. I choose autopartitioning, the installer writes the partition
 table but when it start to write /dev/vtdb0p2 a very cryptic error appears
 about being unable to write - sorry, did not write it down.
 
 The installer then stops.
 
 If I do a fdisk /dev/vtdb0 in the VM I can see the GPT partition being
 there. If I do a fdisk /dev/da2 on the host machine, I can see the GPT
 partition as well, but the VM just doesn't want to write on it.
 
 I even tried changing kern.geom.debugflags=16 as I thought the host machine
 could be locking somehow the drive, but that didn't seem to make any
 difference. I know it was a lame check but I was out of ideas.
 
 So I just wanted to understand if such a scenario is supposed to be
 supported
 
 What I was thinking of, for example, was of having an external iSCSI device
 connected on the hostmachine mapped as a virtual disk for a specific VM, in
 order to speed the VM disk performances.
 
 
 Just another quick question... I have seen some improvements by having the
 VM's virtual disk on ZFS against UFS. Is it just me or is there any real
 improvement by using ZFS?
 
 Thanks a lot.
 
 
 -- *Andrea BrancatelliSchema 31 S.r.l. - Socio UnicoResponsabile ITROMA
 - FIRENZE - PALERMO ITALYTel: +39. 06.98.358.472* *Cell: +39
 331.2488468Fax: +39. 055.71.880.466Societ? del Gruppo SC31 ITALIA*

I'm not answering your question precisely, but can you pass through the
disk controller to the virtual machine instead? I also know that zvol
and iscsi backends work, at least the last time I checked.

- Nikolai Lifanov
___
freebsd-virtualization@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-virtualization
To unsubscribe, send any mail to 
freebsd-virtualization-unsubscr...@freebsd.org