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

[Xen-users] Re: Xen-users Digest, Vol 39, Issue 114


  • To: xen-users@xxxxxxxxxxxxxxxxxxx
  • From: "John @ New Frontiers" <jbanas@xxxxxxxx>
  • Date: Mon, 19 May 2008 14:33:08 -0400
  • Delivery-date: Mon, 19 May 2008 11:33:47 -0700
  • List-id: Xen user discussion <xen-users.lists.xensource.com>


__________
John Banas
Senior Network Administrator

New Frontiers Telecommunications
49 Summit Avenue
Hagerstown, Maryland 21740
P: 301-791-3800

Submit a trouble ticket:
http://www.nfis.com/cgi-bin/troubleticket/ttx.cgi?cmd=newticket

New Frontiers Telecommunications offers local voice and high speed Internet connections for residential and business customers in Western Maryland and West Virginia.


On May 19, 2008, at 2:30 PM, xen-users-request@xxxxxxxxxxxxxxxxxxx wrote:

Send Xen-users mailing list submissions to
        xen-users@xxxxxxxxxxxxxxxxxxx

To subscribe or unsubscribe via the World Wide Web, visit
        http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users
or, via email, send a message with subject or body 'help' to
        xen-users-request@xxxxxxxxxxxxxxxxxxx

You can reach the person managing the list at
        xen-users-owner@xxxxxxxxxxxxxxxxxxx

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Xen-users digest..."


Today's Topics:

  1. Re: vcpus higher than real cpus possible? (Mark Williamson)
  2. Re: Live Migration Query (Asim)
  3. Re: Getting dumU real mem usage (Mark Williamson)
  4. Re: sEDF Scheduling question (Todd Deshane)
  5. Re: Live migration not moving the disk image. Is is        supposed
     to? (Mark Williamson)
  6. Re: IBM x445, anyone using it? (isplist@xxxxxxxxxxxx)
  7. RE: Loading FreeBSD in an HVM-domain on Intel-basedhost    with
     GRUB (Patro, Sumant)


----------------------------------------------------------------------

Message: 1
Date: Mon, 19 May 2008 15:32:15 +0100
From: Mark Williamson <mark.williamson@xxxxxxxxxxxx>
Subject: Re: [Xen-users] vcpus higher than real cpus possible?
To: xen-users@xxxxxxxxxxxxxxxxxxx
Cc: "\[STD\]Ein" <ein@xxxxxxxxxxxxxx>, Stefan de Konink
        <skinkie@xxxxxxxxx>, 'David Hl?cik' <david@xxxxxxxxx>
Message-ID: <200805191532.17778.mark.williamson@xxxxxxxxxxxx>
Content-Type: text/plain;  charset="utf-8"

Having more vcpuÂ?s assigned to a domu than real cpuÂ?s is not a supported
scenario, although it does seem to work.  The scheduling of such a
configuration is a nightmare though, and your guest system will be
interrupting itself when it believes concurrent processing is taking place, which could (will?) lead to instability. You would likely see more cpu
time spent on state switching than actual work in such a setup.

It shouldn't give you instability but it is not efficient so I wouldn't
recommend it unless you need it for some reason.

Think of VCPUs in a VM as like rather like threads in a process. The system can cope with having more of them than there are logical CPUs, it just won't
be able to run them all at once.

One difference is that although there are efficient ways of having more threads in a process than there are logical CPUs, it's very rarely going to be efficient to have more VCPUs than logical CPUs. It misleads the guest OS into thinking it's on a bigger system than it is, resulting in it wasting time trying to use the extra CPUs. As [STD]ein said, the VCPUs will compete
with each other for time on the real machine - wasting time.

For a PV domU you can hot-unplug CPUs from a running domain in order to optimise it for the physical system it's running on. I don't know if this
has been done for an HVM guest, I'd assume you have to set the number
correctly at boot time.

Cheers,
Mark

 _____

From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
[mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of David Hlácik
Sent: May 17, 2008 3:51 PM
To: Stefan de Konink
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] vcpus higher than real cpus possible?

But best approach is to use VCPUs same as real CPUs to achieve best
performace right?
Thanks,

D.
On Sat, May 17, 2008 at 10:32 PM, Stefan de Konink <skinkie@xxxxxxxxx>
wrote:

On Sat, 17 May 2008, [ISO-8859-2] David Hláèik wrote:
Hello, i have just migrated machine to xen server where only 2 cpus are
available. I have copied config from machine where domU was located

before.

That machine had 4 CPUS (QUADCORE), current machine has 2 CPUS (1 XEON
2cores).

I have forgot to change vcpus from 4 to 2 in config, but .. what really surprised me ... machine started . How it is possible? It was really slow
alltought and not possible for working (Windows server 2003).

Since Xen is always doing this :) Think for one processor having 2 DomU's running. Technically you have then 3 vcpu's on the same moment. It is just
about scheduling tasks for a vcpu.


Stefan



--
Push Me Pull You - Distributed SCM tool (http://www.cl.cam.ac.uk/~maw48/pmpu/ )



------------------------------

Message: 2
Date: Mon, 19 May 2008 07:36:41 -0700
From: Asim <linkasim@xxxxxxxxx>
Subject: Re: [Xen-users] Live Migration Query
To: "Stein, Tim" <Tim.O.Stein@xxxxxxxxxxxxxxxxxxx>
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Message-ID:
        <7ef321c10805190736x1a0dadf7w904f54d3a2ce6ef7@xxxxxxxxxxxxxx>
Content-Type: text/plain; charset="iso-8859-1"

I'm using Xen 3.1.4. It is a 32 bit host on a 32 bit kernel for dom0 but the
machine is 64 bit.
Did you get this error as well => ERROR Internal error: Couldn't enable
shadow mode

I thought it has something to do with PAE mode.

Thanks,
Asim




On Sun, May 18, 2008 at 11:18 PM, Stein, Tim <
Tim.O.Stein@xxxxxxxxxxxxxxxxxxx> wrote:

Am Sonntag, 18. Mai 2008 schrieb Asim:

Error: /usr/lib/xen/bin/xc_save 23 3 0 0 1 failed

What version of Xen are you using? I had a similar problem when trying to migrate a 32bit guest under a 64bit host with Xen 3.04. Xen 3.2 solved this
issue for me.

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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://lists.xensource.com/archives/html/xen-users/attachments/20080519/9abf540b/attachment.htm

------------------------------

Message: 3
Date: Mon, 19 May 2008 17:21:52 +0100
From: Mark Williamson <mark.williamson@xxxxxxxxxxxx>
Subject: Re: [Xen-users] Getting dumU real mem usage
To: xen-users@xxxxxxxxxxxxxxxxxxx
Cc: MS <ms@xxxxxxxxxx>
Message-ID: <200805191721.53913.mark.williamson@xxxxxxxxxxxx>
Content-Type: text/plain;  charset="iso-8859-15"

Hi there,

i need to get at my dom0 the real memory usage of a domU.
The real used memoy (like /proc/meminfo), not just the allocated mem who
is shown eg. at xentop.

There's not currently a way to do this. Dan Magenheimer is working on some metrics for determining memory pressure in a domU; I don't know if it's exactly what you're looking for but it's certainly related. His work is aimed at automatically ballooning guests according to what memory they need,
so that the overall throughput of your Xen system can be higher.

Cheers,
mark

--
Push Me Pull You - Distributed SCM tool (http://www.cl.cam.ac.uk/~maw48/pmpu/ )



------------------------------

Message: 4
Date: Mon, 19 May 2008 13:48:21 -0400
From: "Todd Deshane" <deshantm@xxxxxxxxx>
Subject: Re: [Xen-users] sEDF Scheduling question
To: "Koen van Besien" <koen.vanbesien@xxxxxxxxx>
Cc: xen-users <xen-users@xxxxxxxxxxxxxxxxxxx>
Message-ID:
        <1e16a9ed0805191048r28ff3c46rceaa9a7ca1dc08a2@xxxxxxxxxxxxxx>
Content-Type: text/plain; charset=ISO-8859-1

Hi Koen,

On Sun, May 18, 2008 at 8:55 AM, Koen van Besien
<koen.vanbesien@xxxxxxxxx> wrote:
Hi,

I was wondering if there is some more information about the sEDF
scheduler and its parameters.
Or people who have some experience with it.


As I recall, sEDF has been deprecated (replaced with the credit scheduler).

See: http://lists.xensource.com/archives/html/xen-users/2007-01/msg00312.html

Cheers,
Todd



------------------------------

Message: 5
Date: Mon, 19 May 2008 19:11:03 +0100
From: Mark Williamson <mark.williamson@xxxxxxxxxxxx>
Subject: Re: [Xen-users] Live migration not moving the disk image. Is
        is      supposed to?
To: xen-users@xxxxxxxxxxxxxxxxxxx
Cc: 'Stefan de Konink' <skinkie@xxxxxxxxx>,       "Jamie J. Begin"
        <begin@xxxxxxxxxxxxxxx>
Message-ID: <200805191911.06024.mark.williamson@xxxxxxxxxxxx>
Content-Type: text/plain;  charset="iso-8859-1"

From: Stefan de Konink [mailto:skinkie@xxxxxxxxx]
Sent: Thursday, May 15, 2008 7:28 PM
To: Jamie J. Begin
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] Live migration not moving the disk image. Is is
supposed to?

On Thu, 15 May 2008, Jamie J. Begin wrote:
From my reading, it looks like performing "xm migrate -live domU

target_server" ought to also copy the disk image.  Am I right?

When I attempt this procedure, everything looks to work except for the

image

file getting moved. The command takes about 30 seconds to complete and

then

"xm list" shows the domU to be gone from the old host and running on
the

new

host.  The SSH connection to the domU also continues to function
without

a

hiccup. However, the disk image isn't located anywhere on the new

server.

Also, the xen log on the new server lists the following error:

[2008-05-15 19:10:21 xend 3790] ERROR (XendCheckpoint:275) Device 51712
(tap) could not be connected.

/virt/myserver.img does not exist

That sounds pretty much like expected behavior. This is why you require something like NFS, DRBD, iSCSI to have the storage available over more
than one machine at a time.


Stefan

Ok, thanks. I just wanted to make sure that this was expected behavior.
Yeah, iSCSI is forthcoming for the production environment.  I'm just
testing a few things here.

You do need to have the storage accessible from multiple machines in order to support live migration. You can either do this by having dom0 access the storage (e.g. over SAN or network) and export it to the domU as a VBD (I recommend this, it's more transparent) or by setting up the domU to access its storage directly over the network (e.g. using iSCSI root or NFS root).

Xen includes some NBD-aware block scripts to make it easy to handle NBD-based block devices in an automated way. Stefan has, I believe, published some iSCSI (and Netapp filer specific) block scripts on these mailing lists, which
may also prove useful.

Cheers,
Mark

--
Push Me Pull You - Distributed SCM tool (http://www.cl.cam.ac.uk/~maw48/pmpu/ )



------------------------------

Message: 6
Date: Mon, 19 May 2008 13:15:20 -0500
From: "isplist@xxxxxxxxxxxx" <isplist@xxxxxxxxxxxx>
Subject: Re: [Xen-users] IBM x445, anyone using it?
To: xen-users <xen-users@xxxxxxxxxxxxxxxxxxx>
Message-ID: <2008519131520.519815@leena>
Content-Type: text/plain; charset="ISO-8859-1"

 Yep.  The max for a single node x445 is 8 processors, no dual core
 support that I am aware of.

What type of processors should I be looking for which are properly supported then?

 showed up in RHEL and SLES.  Also, to the best of my recollection,
 they were 32bit only boxes.  And, as I said before, there is no
 hardware virutalization support on those processors or BIOS.

Yes, 32bit.

Just from the little bit you have shared, your problem has NOTHING to
 do with hardware virtualization, unless you are trying to take a HVM
 guest from one Xen box that supports VT, and trying to run it on a
 different Xen box that does NOT support VT.

It's just a fresh CentOS install with Virtualization because I wanted to try out Xen.

 Sounds more like a filesystem issue or something like that.

I'm not sure that there is a problem, perhaps a little configuration which is not installed by default or such?

 Really, use virt-manager or YaST and install a paravirtualized guest
 just to see if it works.  Besides, you have a running system, and a
 running Dom0, so Xen works.  Dom0 is, in reality, just a
 paravirtualized guest OS running on the hypervisor.

That's exactly what I've tried and received the path error :).
Using Virtual Machine Manager, I've tried to install a pre- configured ISO. It was simply what I had handy. I can try something else. Is there something I can try from the command line? I just want a basic Linux install, no servers, nothing for now.

 What does this mean?  You copied a working guest filesystem image to
 your new Xen box, or you copied an install tree?  What OS are you
using for the Host? What OS are you using for the guest? How is your
 install tree set up (file tree, ISO, or what?)

It's an ISO of a pre-installed CentOS OS with qmail pre-configured along with other tools. Just had it handy when I first gave this a quick try.

Again, as I said before, you will not have Fully Virtualized abilities on the x445, it does not support the hardware virtualization code, and
 I am almost sure that it won't even support the processors that have
 HVM code on them (Though that is just a guess, so YMMV there).

That's fine, like I too said, I just happen to have these machines sitting here so was hoping I might be able to use them with Xen. From what I can tell, I won't be needing fully virtualized machines anyhow. I don't tend to consolidate things which I try to get the most power from. In other words, I won't be virtualizing machines which need to be as powerful as they can be. I will be virtualizing machines which don't do a lot, DNS servers for example are pretty simple I/O machines, very low use web, mail, other servers, things like that.

 And to answer your other questions... virtualbox may work, but it's
 rather new, and I've never tried it myself.  VMWare MAY work,

I didn't ask about this but I think it's in the thread somewhere. Either way, if I can use the x445 for non fully virtualized machines, that still would be cool. It would allow me to shut down a couple dozen low use boxes, virtualizing them along with their storage on SAN.

 But again, you have Xen booted, and a running Dom0, so Xen is doing
 what it's supposed to do... my money is on a bad install tree (which
 is what that error you shared seems to point to) or trying to run a
 full virt guest from one Xen box on a second Xen box that does not
 support full virt.

I'll look around for a guest which is known to easily work para- virtualized. If you know of one, please let me know in the meantime. Thanks.

Mike




------------------------------

Message: 7
Date: Mon, 19 May 2008 12:30:01 -0600
From: "Patro, Sumant" <Sumant.Patro@xxxxxxx>
Subject: RE: [Xen-users] Loading FreeBSD in an HVM-domain on
        Intel-basedhost with GRUB
To: "Boris Derzhavets" <bderzhavets@xxxxxxxxx>
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Message-ID:
        <0631C836DBF79F42B5A60C8C8D4E82290109C112@xxxxxxxxxxxxxxxxxxx>
Content-Type: text/plain; charset="us-ascii"

Thank you Boris.





I could install FreeBSD 7.0 DomU with the instructions listed in the
link you provided.



Regards,



Sumant



________________________________

From: Boris Derzhavets [mailto:bderzhavets@xxxxxxxxx]
Sent: Thursday, May 15, 2008 2:15 AM
To: Patro, Sumant
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] Loading FreeBSD in an HVM-domain on
Intel-basedhost with GRUB




Install HVM FreeBSD 7.0 DomU (32-bit) at Xen 3.2.1 CentOS 5.1 Dom0
(64-bit)


http://bderzhavets.blogspot.com/2008/04/install-hvm-freebsd-7.html

----- Original Message ----
From: "Patro, Sumant" <Sumant.Patro@xxxxxxx>
To: Igor Chubin <igor@xxxxxxx>; Keir.Fraser@xxxxxxxxxxxx;
kip.macy@xxxxxxxxx
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Sent: Wednesday, May 14, 2008 10:07:59 PM
Subject: RE: [Xen-users] Loading FreeBSD in an HVM-domain on
Intel-basedhost with GRUB

Hello All,

   I am trying to install FreeBSD 7.0 (8.0-current) as a guest OS
with RHEL 5.1/5.2. I compiled xen 3.2 with vmxassist=n as suggested in
bug 622. After I installed and booted this xen kernel I have difficulty
in starting the VMManager to create guest OS.

   Could anyone please provide details on which Linux OS & the
tools I may use to install FreeBSD as guest.

   I also tried with the instructions @
http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/virtualization
-guest.html with kernel7.0-Release and kernel8.0-current. With both of
them I get a crash on doing xm create xmexample1.bsd -c. The trace is
below :

(XEN) Unhandled page fault in domain 1 on VCPU 0 (ec=0002)
       (XEN) Pagetable walk from 00000472:
       (XEN)  L2[0x000] = 00000000 ffffffff
       (XEN) domain_crash_sync called from entry.S (ff157619)
       (XEN) Domain 1 (vcpu#0) crashed on cpu#1:
       (XEN) ----[ Xen-3.0.4-1  x86_32  debug=n  Not tainted
]----
       (XEN) CPU:    1
       (XEN) EIP:    e019:[<c0458a30>]
       (XEN) EFLAGS: 00000202  CONTEXT: guest
       b  cr4: 000026d0  cr3: 0d873000  cr2: 00000472
       (XEN) ds: e021  es: e021  fs: e021  gs: e021  ss:
e021  cs: e019
       (XEN) Guest stack trace from esp=c0c4fff0:
       (XEN)    00000002 c0458a30 0001e019 00010002
       (XEN) Unhandled page fault in domain 2 on VCPU 0
(ec=0002)
       (XEN) Pagetable walk from 00000472:
       (XEN)  L2[0x000] = 00000000 ffffffff
       (XEN) domain_crash_sync called from entry.S (ff157619)
       (XEN) Domain 2 (vcpu#0) crashed on cpu#1:
       (XEN) ----[ Xen-3.0.4-1  x86_32  debug=n  Not tainted
]----
       (XEN) CPU:    1
       (XEN) EIP:    e019:[<c0458a30>]
       (XEN) EFLAGS: 00000202  CONTEXT: guest
       (XEN) eax: 00000000  ebx: 00000000  ecx: 00000000
edx: 00000000
       (XEN) esi: c0c47000  edi: 00000000  ebp: 00000000
esp: c0c4fff0
       (XEN) cr0: 8005003b  cr4: 000026d0002 c0458a30 0001e019
00010002
       Unhandled page fault in domain 2llllllINIT: Switching to
runlevel: 0


   Thanks in advance. Your help is highly appreciated.

Regards,

Sumant

-----Original Message-----
From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
[mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Igor Chubin
Sent: Thursday, February 07, 2008 10:37 AM
To: Keir.Fraser@xxxxxxxxxxxx
Cc: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] Loading FreeBSD in an HVM-domain on
Intel-basedhost with GRUB



The problem is solved.

Keir Fraser have said in Xen-devel mailing list
that the problem should not exist in the current xen-unstable
(since yesterday).

I have just built xen-unstable and really there is no problem
with usage of /boot/loader in an HVM domain even on a Intel-based host.
So bug 622 does not exist anymore (at least for me).

Thank all and thank Keir again.

On Do, Feb 07, 2008 at 11:32:28 +0100, Oskar Lindgren wrote:
Hi,

Yes, it is possible. That is how I do it. But io under freebsd is
slow..

/ Oskar

Igor Chubin skrev:
Hello all,

please tell me,


is it possible (at least theoretically) to circumvent bug 622 [1] by
using
GRUB instead of /boot/loader to load FreeBSD kernel?

Or it is silly idea: even if we bypass /boot/loader we will not be
able
to run FreeBSD kernel in a HVM-domain on a Intel-based host for some
other reason?


As far as I know, GRUB can't load FreeBSD kernel
directly, but only can make chainloading using /boot/loader.
Certainly, it is not the solution, but there were patches
to GRUB that allow to boot FreeBSD directly [2].

I want to try to use it and eliminate /boot/loader
from boot process at all.


What can you say about this idea?



[1] http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=622
[2] http://savannah.gnu.org/bugs/?10728




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

--
WBR, i.m.chubin


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

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



-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://lists.xensource.com/archives/html/xen-users/attachments/20080519/0298f256/attachment.html

------------------------------

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


End of Xen-users Digest, Vol 39, Issue 114
******************************************


_______________________________________________
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®.