Xen users - Oct 2010 - any opinions on debian vs. opensuse for Xen?

Hi Folks,

I''ve been getting just a little fed up with the state of Xen on Debian 
Lenny - there are a couple of known bugs that lead to periodic kernel 
panics - but fixes haven''t made it into the Lenny distribution. 
I''m
sort of waiting to see how things shake out with Debian Squeeze, when it 
becomes stable, but...

I''ve been noticing that Suse (both the enterprise and OpenSuse
versions)
seem to stress support for Xen, as well as for BTRFS (which I''ve been 
keeping my eyes on).

So... I''m wondering if anybody has any opinions, based on direct, 
hands-on experience on Debian vs. OpenSuse as a base for a configuration 
consisting of 2 production machines configured for high-availability 
failover of Xen VMs, using DRBD and Pacemaker?

Thanks much,

Miles Fidelman

-- 
In theory, there is no difference between theory and practice.
In<fnord>  practice, there is.   .... Yogi Berra



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

In my tests opensuse works with Xen so perfectly so I become afraid it.

It works even more better, than RHEL, which one have few minor glitches
in memory usage under xen.

В Сбт, 16/10/2010 в 14:27 -0400, Miles Fidelman пишет:
> Hi Folks,
> 
> I''ve been getting just a little fed up with the state of Xen on
Debian
> Lenny - there are a couple of known bugs that lead to periodic kernel 
> panics - but fixes haven''t made it into the Lenny distribution. 
I''m
> sort of waiting to see how things shake out with Debian Squeeze, when it 
> becomes stable, but...
> 
> I''ve been noticing that Suse (both the enterprise and OpenSuse
versions)
> seem to stress support for Xen, as well as for BTRFS (which I''ve
been
> keeping my eyes on).
> 
> So... I''m wondering if anybody has any opinions, based on direct, 
> hands-on experience on Debian vs. OpenSuse as a base for a configuration 
> consisting of 2 production machines configured for high-availability 
> failover of Xen VMs, using DRBD and Pacemaker?
> 
> Thanks much,
> 
> Miles Fidelman
> 


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

Avoid any debian OS as a Dom0 like the plague. In my experience, it has 
lead to nothing but problems (lockups and panics).

I personally use CentOS. I''ve also heard good things about OpenSuse and
Xen...


On 17/10/10 00:37, George Shuklin wrote:
> In my tests opensuse works with Xen so perfectly so I become afraid it.
>
> It works even more better, than RHEL, which one have few minor glitches
> in memory usage under xen.
>
> В Сбт, 16/10/2010 в 14:27 -0400, Miles Fidelman пишет:
>    
>> Hi Folks,
>>
>> I''ve been getting just a little fed up with the state of Xen
on Debian
>> Lenny - there are a couple of known bugs that lead to periodic kernel
>> panics - but fixes haven''t made it into the Lenny
distribution.  I''m
>> sort of waiting to see how things shake out with Debian Squeeze, when
it
>> becomes stable, but...
>>
>> I''ve been noticing that Suse (both the enterprise and OpenSuse
versions)
>> seem to stress support for Xen, as well as for BTRFS (which
I''ve been
>> keeping my eyes on).
>>
>> So... I''m wondering if anybody has any opinions, based on
direct,
>> hands-on experience on Debian vs. OpenSuse as a base for a
configuration
>> consisting of 2 production machines configured for high-availability
>> failover of Xen VMs, using DRBD and Pacemaker?
>>
>> Thanks much,
>>
>> Miles Fidelman
>>
>>      
>
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
>    
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On 10/16/2010 01:27 PM, Miles Fidelman wrote:
> Hi Folks,
>
> I''ve been getting just a little fed up with the state of Xen on
Debian
> Lenny - there are a couple of known bugs that lead to periodic kernel
> panics - but fixes haven''t made it into the Lenny distribution.
I''m sort
> of waiting to see how things shake out with Debian Squeeze, when it
> becomes stable, but...
>
> I''ve been noticing that Suse (both the enterprise and OpenSuse
versions)
> seem to stress support for Xen, as well as for BTRFS (which I''ve
been
> keeping my eyes on).
>
> So... I''m wondering if anybody has any opinions, based on direct,
> hands-on experience on Debian vs. OpenSuse as a base for a configuration
> consisting of 2 production machines configured for high-availability
> failover of Xen VMs, using DRBD and Pacemaker?
>
> Thanks much,
>
> Miles Fidelman
>
I have been using lenny for over a year now. I had some lockups at first 
but I never could tell if it was just on vm that was crashing and 
bringing the whole setup down. Since I have done cpu pinning and better 
ram management, I haven''t had any problems.
I have been using squeeze as a desktop for over a year now with no 
problems. The only problems right now in squeeze will probably be 
related to installer quirks. I would give it a try. I am going to 
upgrade my set up probably in Dec. or so.

Sam

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

Sam wrote:
>I have been using lenny for over a year now. I had some lockups at 
>first but I never could tell if it was just on vm that was crashing 
>and bringing the whole setup down. Since I have done cpu pinning and 
>better ram management, I haven''t had any problems.
>I have been using squeeze as a desktop for over a year now with no 
>problems. The only problems right now in squeeze will probably be 
>related to installer quirks. I would give it a try. I am going to 
>upgrade my set up probably in Dec. or so.
I''ve also running Lenny (with a 2.6.18 kernel for Dom0) on three 
boxes - apart from one issue*, it "just runs" and I''ve never
had a
machine freeze, lockup, or crash.

* The well known xenconsoled bug that makes a guest stop while 
consuming about 100% of a cpu core. Restarting xenconsoled will fix 
it and allow the guest to continue from where it was. It''s not that 
common, I''d estimate less than once a month between 3 hosts running 
about 15 guests between them.

I''ve been keeping an eye on Squeeze and Lenny Backports. When I can 
get another box I''ll have a look at one or the other.

-- 
Simon Hobson

Visit http://www.magpiesnestpublishing.co.uk/ for books by acclaimed
author Gladys Hobson. Novels - poetry - short stories - ideal as
Christmas stocking fillers. Some available as e-books.

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

Sam wrote:
> On 10/16/2010 01:27 PM, Miles Fidelman wrote:
>> I''ve been getting just a little fed up with the state of Xen
on Debian
>> Lenny - there are a couple of known bugs that lead to periodic kernel
>> panics - but fixes haven''t made it into the Lenny
distribution. I''m sort
>> of waiting to see how things shake out with Debian Squeeze, when it
>> becomes stable, but...
>
> I have been using lenny for over a year now. I had some lockups at 
> first but I never could tell if it was just on vm that was crashing 
> and bringing the whole setup down. Since I have done cpu pinning and 
> better ram management, I haven''t had any problems.
Based on the console errors I''ve seen when I have a kernel panic, and 
some of the open bug reports I''ve seen, I''m guessing that cpu
pinning
would solve my immediate problem.

But I''m not sure what an effective configuration would be - my current 
environment has 2 CPUs and 5 VMs (dom0 + 4 domUs) per server.  Not sure 
how to set up the pinning, and I expect there would be a performance 
hit.  So much easier to let the hypervisor allocate CPUs dynamically - 
if it worked right.

Sam... could you say a bit more about the configuration details of what 
you''re doing re. "cpu pinning and better ram management?"

Thanks,

Miles

-- 
In theory, there is no difference between theory and practice.
In<fnord>  practice, there is.   .... Yogi Berra



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

On 10/17/2010 08:14 AM, Miles Fidelman wrote:
> Sam wrote:
>> On 10/16/2010 01:27 PM, Miles Fidelman wrote:
>>> I''ve been getting just a little fed up with the state of
Xen on Debian
>>> Lenny - there are a couple of known bugs that lead to periodic
kernel
>>> panics - but fixes haven''t made it into the Lenny
distribution. I''m sort
>>> of waiting to see how things shake out with Debian Squeeze, when it
>>> becomes stable, but...
>>
>> I have been using lenny for over a year now. I had some lockups at
>> first but I never could tell if it was just on vm that was crashing
>> and bringing the whole setup down. Since I have done cpu pinning and
>> better ram management, I haven''t had any problems.
> Based on the console errors I''ve seen when I have a kernel panic,
and
> some of the open bug reports I''ve seen, I''m guessing that
cpu pinning
> would solve my immediate problem.
>
> But I''m not sure what an effective configuration would be - my
current
> environment has 2 CPUs and 5 VMs (dom0 + 4 domUs) per server. Not sure
> how to set up the pinning, and I expect there would be a performance
> hit. So much easier to let the hypervisor allocate CPUs dynamically - if
> it worked right.
>
> Sam... could you say a bit more about the configuration details of what
> you''re doing re. "cpu pinning and better ram
management?"
>
> Thanks,
>
> Miles
>
I am running on a dual 1.6ghz cpu intel atom motherboard so of course 
this is less than ideal. The atom cpu, even though 2 core, has hyper 
threading which makes dom0 think it has 4 cpus so you have to do some 
testing to figure out which cpu''s are actually on a separate core. dom0
by default has access to all cores. I only have 2 domU''s. One has:

cpus = "1,3"
vcpus = 2
memory = 732

while the other has:
cpus = "0,2"
vcpus = 2
memory = 732

Again, note that even though they were passed 2 cpu''s, they are just 
actually getting one hyperthreaded cpu core. dom0 also additionally gets 
a set amount of ram at boot time.

 From menu.lst:
title           Xen 3.2-1-amd64 / Debian GNU/Linux, kernel 
2.6.26-2-xen-amd64
root            (hd0,0)
kernel          /xen-3.2-1-amd64.gz dom0_mem=256M
module          /vmlinuz-2.6.26-2-xen-amd64 root=/dev/mapper/vg0-xen ro 
console=tty0
module          /initrd.img-2.6.26-2-xen-amd64


That is not much ram for dom0, but not much is going on there in my set 
up. / is only taking up 900mb. After 4 months of uptime, it finally 
started using some swap (a whole 10mb). In the /etc/rc.local of dom0 I 
passed "xm sched-credit -d Domain-0 -w 512" to give dom0 higher
priority
over other vm''s also.

So in your case I would figure out if any of your domU''s are considered
"more stable" than the others. Put the domU''s that you trust
stability
wise on on cpu, the others on the other cpu. Of course unless they are 
all equal or have the same distro, then just just put 2 on one cpu and 2 
on the other. dom0 will have access to both cpus. I doubt 2 domU''s on 
the same core would crash at the same time and and use up all the 
resources. Nothing should be crashing anyway.

I would also go with debian squeeze unless you are having trouble with 
the debian installer. I normally use the net install disk 
http://www.debian.org/distrib/netinst and then use aptitude to install 
the rest of my packages. I don''t ever use the preset tasks to install 
any thing.

The wiki has some tips: http://wiki.debian.org/Xen

To install the domU''s use this:
lenny:
http://ftp.debian.org/debian/dists/lenny/main/installer-i386/current/images/netboot/xen/xm-debian.cfg
sqeeze:
http://ftp.debian.org/debian/dists/squeeze/main/installer-amd64/current/images/netboot/xen/xm-debian.cfg

Of course, all of these debian ways I have showed you only involve the 
console and no gui installers. If you are wanting that, you might have 
to look else where.

Sam

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

Il 16/10/2010 20:27, Miles Fidelman ha scritto:
> Hi Folks,
> 
> I''ve been getting just a little fed up with the state of Xen on
Debian
> Lenny - there are a couple of known bugs that lead to periodic kernel
> panics - but fixes haven''t made it into the Lenny distribution. 
I''m
> sort of waiting to see how things shake out with Debian Squeeze, when it
> becomes stable, but...
> 
> I''ve been noticing that Suse (both the enterprise and OpenSuse
versions)
> seem to stress support for Xen, as well as for BTRFS (which I''ve
been
> keeping my eyes on).
> 
> So... I''m wondering if anybody has any opinions, based on direct,
> hands-on experience on Debian vs. OpenSuse as a base for a configuration
> consisting of 2 production machines configured for high-availability
> failover of Xen VMs, using DRBD and Pacemaker?
Hi,
IMHO suse (in particular SLES) is ready for production.
I''m actually testing it on some supermicro servers based on mobo
H8DGU-LN4F+ and 2x AMD 6168 and it works very well.

When you look to something like a virtualization server you have to
consider many aspects:
- how do you managed storage for the VMs
- what kind network between nodes
- hypervisor features
and so on.

Actually SUSE sles or opensuse have a recent kernel with stable patches,
a working cluster suite (openais, corosync, heartbeat, pacemaker), a
clusterized volume managed CLVM, two clusterized filesystems OCFS2 and GFS.

...and of course suse have a good infiniband stack and it''s supported
by
many "complementary" products ...for example convirture.

I''m investigating the possibility to port some of the above to a
different distro (in particular gentoo) because I want to creare a
micro-image served via PXE for my nodes ...but actually I''m in an early
stage ...so I don''t know when something usable will be available.


-- 
Christian Zoffoli (XMerlin)

You never change things by fighting the existing reality. To change
something, build a new model that makes the existing model obsolete."
-- R. Buckminster Fuller

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

Sam,

Thanks for the details!  And...

Sam wrote:
> I am running on a dual 1.6ghz cpu intel atom motherboard so of course 
> this is less than ideal. The atom cpu, even though 2 core, has hyper 
> threading which makes dom0 think it has 4 cpus so you have to do some 
> testing to figure out which cpu''s are actually on a separate core.
> dom0 by default has access to all cores. I only have 2 domU''s. One
has:
I''m not sure if my system has hyperthreading or not - I guess
I''m going
to have to figure that out first.

Do you know if there''s a way to pin dom0 to a specific core - I think 
I''m seeing kernel panics caused by dom0 (SMP errors that show up every 
once in a while).  I''m running a high-availability configuration
running
Xen over DRBD over software RAID10 - and I expect panics occur when all 
the bits line up just right before swapping into/out of dom0 - but I 
don''t have the instrumentation to drill down that
far.
> I would also go with debian squeeze unless you are having trouble with 
> the debian installer. I normally use the net install disk 
> http://www.debian.org/distrib/netinst and then use aptitude to install 
> the rest of my packages. I don''t ever use the preset tasks to
install
> any thing.
I think I might just try going to Squeeze.  As I read the bug reports, 
the ones that seem to reference the kernal panics I''m seeing are shown 
as closed in Squeeze.

Are you running Xen 3 or 4?

Thanks,

Miles

-- 
In theory, there is no difference between theory and practice.
In<fnord>  practice, there is.   .... Yogi Berra



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

Christian Zoffoli wrote:
> Hi,
> IMHO suse (in particular SLES) is ready for production.
> I''m actually testing it on some supermicro servers based on mobo
> H8DGU-LN4F+ and 2x AMD 6168 and it works very well.
>
>
>    
In my case, I''m using slightly older hardware - Supermicro P8SCT 
motherboard, P4 CPU.  If I had something more recent (i.e., with 
hardware virtualization support), I''d probably be exploring KVM - seems
a bit better supported than Xen these days.

Thanks for the feedback re. suse.

Miles

-- 
In theory, there is no difference between theory and practice.
In<fnord>  practice, there is.   .... Yogi Berra



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

On 10/17/2010 02:59 PM, Miles Fidelman wrote:
> Sam,
>
> Thanks for the details! And...
>
> Sam wrote:
>> I am running on a dual 1.6ghz cpu intel atom motherboard so of course
>> this is less than ideal. The atom cpu, even though 2 core, has hyper
>> threading which makes dom0 think it has 4 cpus so you have to do some
>> testing to figure out which cpu''s are actually on a separate
core.
>> dom0 by default has access to all cores. I only have 2 domU''s.
One has:
> I''m not sure if my system has hyperthreading or not - I guess
I''m going
> to have to figure that out first.
>
> Do you know if there''s a way to pin dom0 to a specific core - I
think
> I''m seeing kernel panics caused by dom0 (SMP errors that show up
every
> once in a while). I''m running a high-availability configuration
running
> Xen over DRBD over software RAID10 - and I expect panics occur when all
> the bits line up just right before swapping into/out of dom0 - but I
> don''t have the instrumentation to drill down that far.
>> I would also go with debian squeeze unless you are having trouble with
>> the debian installer. I normally use the net install disk
>> http://www.debian.org/distrib/netinst and then use aptitude to install
>> the rest of my packages. I don''t ever use the preset tasks to
install
>> any thing.
> I think I might just try going to Squeeze. As I read the bug reports,
> the ones that seem to reference the kernal panics I''m seeing are
shown
> as closed in Squeeze.
>
> Are you running Xen 3 or 4?
>
> Thanks,
>
> Miles
>
Running 3.2. You can pin with the cpu settings (dom0-cpus) in 
xend-config.sxp

Sam

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

Sam wrote:
> Running 3.2. You can pin with the cpu settings (dom0-cpus) in 
> xend-config.sxp
Thanks!

In theory, there is no difference between theory and practice.
In<fnord>  practice, there is.   .... Yogi Berra



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