Netbsd XEN nothing I can boot in

On Fri, Sep 15, 2006 at 02:46:11PM +1200, Ivan Vari wrote:
> >I'm not sure it's NetBSD, or if it's an issue
with the way this box reports
> >its memory and Xen. The NetBSD kernel trusts what
Xen tells him about 
> >useable
> >memory.
> 
> Thanks for the explanation Manuel. Some more problems
though...
> I have my first 3.1RC2 candidate running, nothing
special. Strictly 
> minimal install, no games, X11 on 10GB LVM for xbd0a.
Everything is on 
> separated slice inside of the block device 1GB /home,
4GB /usr, 512M /, 
> 2Gb /var, 2GB /tmp,  1GB swap and 64M for mfs.
> 
> I followed the howto from wiki.xensource.com and copied
across the xbd* 
> and rxbd* devices after install as suggested.

Not sure what you mean here

> 
> Now after I boot the VM in (using the netbsd-XEN3_DOMU
kernel now) I can 
> log in but the VM stops responding after round about 5
minutes. Whatever 
> I do it happens after 5 min no excuse.
> 
> On another console logged into the dom0 xm top shows
that the allocated 
> 512M is being chewed up, the CPU is 100%. Because of
this my desktop is 
> also becoming non responsive and starts swapping like
nuts. Obviously I 
> log into the dom0 via ssh and start the VM from there
but it should not 
> be an issue I suppose.

I don't understant what you mean here. Is the Xen box you
desktop too ?
Otherwise how could a ssh session make your desktop swap ?
When you says the CPU is 100%, is it dom0's CPU or domU's
CPU ?

Hum, someone on this list posted a note about a similar
issue; if the dom0
CPU is at 100% the domU can hang. This is a property of the
default
Xen's scheduler parameters. Changing this to have a non-0
minimum time slice
for domUs helps.

-- 
Manuel Bouyer, LIP6, Universite Paris VI.          
Manuel.Bouyerlip6.fr
     NetBSD: 26 ans d'experience feront toujours la
difference
--

Sorry for the confusions...

> Not sure what you mean here
http://w
iki.xensource.com/xenwiki/NetBSDdomU

Before the first reboot there's a suggestion to:
mount /dev/xbd0a /mnt
cd /mnt/dev
cp -pR /dev/rxbd* .
cp -pR /dev/xbd* .
halt -p

It was about an older NetBSD 2 domU install but anyway I
have actually
tried without this however same hanging occur.

> I don't understant what you mean here. Is the Xen box
you desktop too ?
> Otherwise how could a ssh session make your desktop
swap ?
> When you says the CPU is 100%, is it dom0's CPU or
domU's CPU ?

No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
From the ssh console running on the dom0 I start up the VM
as root with -c
paramter of course then I log into the VM as root. 5 minutes
later it
stops responding. In the meantime I have another dom0 ssh
connection and
that shows the CPU load to be 100% and the fully used 512M
memory by xm
top.

When this hanging happens it drives my desktop Fedora nuts.
It becomes non
responsive too and start swapping... I didn't understand
this neither but
let's not worry about my dekstop. 

> Xen's scheduler parameters. Changing this to have a
non-0 minimum time
> slice
 Could you point me to the right place about it? I don't
understand this.

I really appreciate your efforts.

Thanks
Ivan

Sorry for the confusions...

> Not sure what you mean here
http://w
iki.xensource.com/xenwiki/NetBSDdomU

Before the first reboot there's a suggestion to:
mount /dev/xbd0a /mnt
cd /mnt/dev
cp -pR /dev/rxbd* .
cp -pR /dev/xbd* .
halt -p

It was about an older NetBSD 2 domU install but anyway I
have actually
tried without this however same hanging occur.

> I don't understant what you mean here. Is the Xen box
you desktop too ?
> Otherwise how could a ssh session make your desktop
swap ?
> When you says the CPU is 100%, is it dom0's CPU or
domU's CPU ?

No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
From the ssh console running on the dom0 I start up the VM
as root with -c
paramter of course then I log into the VM as root. 5 minutes
later it
stops responding. In the meantime I have another dom0 ssh
connection and
that shows the CPU load to be 100% and the fully used 512M
memory by xm
top.

When this hanging happens it drives my desktop Fedora nuts.
It becomes non
responsive too and start swapping... I didn't understand
this neither but
let's not worry about my dekstop. 

> Xen's scheduler parameters. Changing this to have a
non-0 minimum time
> slice
 Could you point me to the right place about it? I don't
understand this.

I really appreciate your efforts.

Thanks
Ivan

Sorry for the confusions...

> Not sure what you mean here
http://w
iki.xensource.com/xenwiki/NetBSDdomU

Before the first reboot there's a suggestion to:
mount /dev/xbd0a /mnt
cd /mnt/dev
cp -pR /dev/rxbd* .
cp -pR /dev/xbd* .
halt -p

It was about an older NetBSD 2 domU install but anyway I
have actually
tried without this however same hanging occur.

> I don't understant what you mean here. Is the Xen box
you desktop too ?
> Otherwise how could a ssh session make your desktop
swap ?
> When you says the CPU is 100%, is it dom0's CPU or
domU's CPU ?

No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
From the ssh console running on the dom0 I start up the VM
as root with -c
paramter of course then I log into the VM as root. 5 minutes
later it
stops responding. In the meantime I have another dom0 ssh
connection and
that shows the CPU load to be 100% and the fully used 512M
memory by xm
top.

When this hanging happens it drives my desktop Fedora nuts.
It becomes non
responsive too and start swapping... I didn't understand
this neither but
let's not worry about my dekstop. 

> Xen's scheduler parameters. Changing this to have a
non-0 minimum time
> slice
 Could you point me to the right place about it? I don't
understand this.

I really appreciate your efforts.

Thanks
Ivan

Sorry for the confusions...

> Not sure what you mean here
http://w
iki.xensource.com/xenwiki/NetBSDdomU

Before the first reboot there's a suggestion to:
mount /dev/xbd0a /mnt
cd /mnt/dev
cp -pR /dev/rxbd* .
cp -pR /dev/xbd* .
halt -p

It was about an older NetBSD 2 domU install but anyway I
have actually
tried without this however same hanging occur.

> I don't understant what you mean here. Is the Xen box
you desktop too ?
> Otherwise how could a ssh session make your desktop
swap ?
> When you says the CPU is 100%, is it dom0's CPU or
domU's CPU ?

No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
From the ssh console running on the dom0 I start up the VM
as root with -c
paramter of course then I log into the VM as root. 5 minutes
later it
stops responding. In the meantime I have another dom0 ssh
connection and
that shows the CPU load to be 100% and the fully used 512M
memory by xm
top.

When this hanging happens it drives my desktop Fedora nuts.
It becomes non
responsive too and start swapping... I didn't understand
this neither but
let's not worry about my dekstop. 

> Xen's scheduler parameters. Changing this to have a
non-0 minimum time
> slice
 Could you point me to the right place about it? I don't
understand this.

I really appreciate your efforts.

Thanks
Ivan

On Fri, Sep 15, 2006 at 10:39:19PM +1200, variianiwa.co.nz wrote:
> Sorry for the confusions...
> 
> > Not sure what you mean here
> http://w
iki.xensource.com/xenwiki/NetBSDdomU
> 
> Before the first reboot there's a suggestion to:
> mount /dev/xbd0a /mnt
> cd /mnt/dev
> cp -pR /dev/rxbd* .
> cp -pR /dev/xbd* .
> halt -p
> 
> It was about an older NetBSD 2 domU install but anyway
I have actually
> tried without this however same hanging occur.

It's not needed any more

> 
> > I don't understant what you mean here. Is the Xen
box you desktop too ?
> > Otherwise how could a ssh session make your
desktop swap ?
> > When you says the CPU is 100%, is it dom0's CPU
or domU's CPU ?
> 
> No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
> >From the ssh console running on the dom0 I start up
the VM as root with -c
> paramter of course then I log into the VM as root. 5
minutes later it
> stops responding. In the meantime I have another dom0
ssh connection and
> that shows the CPU load to be 100% and the fully used
512M memory by xm
> top.

Who is at 100%, dom0 or the domU ? The fast that the 512M
are all used
may not be a problem.

> 
> When this hanging happens it drives my desktop Fedora
nuts. It becomes non
> responsive too and start swapping... I didn't
understand this neither but
> let's not worry about my dekstop. 

It would give a hint about what the dom0 or domU is doing,
though. A top
on your desktop should show what process is becoming crasy,
isn't it ?

> 
> > Xen's scheduler parameters. Changing this to have
a non-0 minimum time
> > slice
>  Could you point me to the right place about it? I
don't understand this.

I never played with this either; but I think it's with
xm sched-sedf <domain> -s 
you can try
xm sched-sedf <domain> -s 10

-- 
Manuel Bouyer <bouyerantioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la
difference
--

On Fri, Sep 15, 2006 at 10:39:19PM +1200, variianiwa.co.nz wrote:
> Sorry for the confusions...
> 
> > Not sure what you mean here
> http://w
iki.xensource.com/xenwiki/NetBSDdomU
> 
> Before the first reboot there's a suggestion to:
> mount /dev/xbd0a /mnt
> cd /mnt/dev
> cp -pR /dev/rxbd* .
> cp -pR /dev/xbd* .
> halt -p
> 
> It was about an older NetBSD 2 domU install but anyway
I have actually
> tried without this however same hanging occur.

It's not needed any more

> 
> > I don't understant what you mean here. Is the Xen
box you desktop too ?
> > Otherwise how could a ssh session make your
desktop swap ?
> > When you says the CPU is 100%, is it dom0's CPU
or domU's CPU ?
> 
> No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
> >From the ssh console running on the dom0 I start up
the VM as root with -c
> paramter of course then I log into the VM as root. 5
minutes later it
> stops responding. In the meantime I have another dom0
ssh connection and
> that shows the CPU load to be 100% and the fully used
512M memory by xm
> top.

Who is at 100%, dom0 or the domU ? The fast that the 512M
are all used
may not be a problem.

> 
> When this hanging happens it drives my desktop Fedora
nuts. It becomes non
> responsive too and start swapping... I didn't
understand this neither but
> let's not worry about my dekstop. 

It would give a hint about what the dom0 or domU is doing,
though. A top
on your desktop should show what process is becoming crasy,
isn't it ?

> 
> > Xen's scheduler parameters. Changing this to have
a non-0 minimum time
> > slice
>  Could you point me to the right place about it? I
don't understand this.

I never played with this either; but I think it's with
xm sched-sedf <domain> -s 
you can try
xm sched-sedf <domain> -s 10

-- 
Manuel Bouyer <bouyerantioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la
difference
--

> Who is at 100%, dom0 or the domU ? The fast that the
512M are all used
> may not be a problem.

DomU NetBSD is. Something must be going wrong and that is my
main concern.
A plain (minimal nx100MB) system and a login shouldn't use
100% CPU
whatever it is (server, 486SX or VM) I reckon but I am
perhaps wrong. I
didn't have a chance to configure rc.conf at all so no
service is running
whatsoever apart from the sendmail and perhaps inetd.

Just for a test I am gonna disable SMP and HT too. (I have
set them back
since others suggested that its ok for dom0 and my problem
was the memory
anyway) Also I will try linux domU but that is not my goal
so gotta get
BSD going 


> It would give a hint about what the dom0 or domU is
doing, though. A top
> on your desktop should show what process is becoming
crasy, isn't it ?

Okay, will investigate that as well.

> I never played with this either; but I think it's with
> xm sched-sedf <domain> -s
> you can try
> xm sched-sedf <domain> -s 10

Ok I will try this too but rather fix the root cause... If
it's not needed
for most cases I wouldn't be happy to use it. It would be
perhaps just a
bypass...

Thanks
Ivan

On Sat, Sep 16, 2006 at 07:38:25AM +1200, variianiwa.co.nz wrote:
> > Who is at 100%, dom0 or the domU ? The fast that
the 512M are all used
> > may not be a problem.
> 
> DomU NetBSD is. Something must be going wrong and that
is my main concern.
> A plain (minimal nx100MB) system and a login shouldn't
use 100% CPU
> whatever it is (server, 486SX or VM) I reckon but I am
perhaps wrong. I
> didn't have a chance to configure rc.conf at all so no
service is running
> whatsoever apart from the sendmail and perhaps inetd.

Yes, something is wrong here. I suspect looping in kernel.
You can try
entering ddb (+++++) and see where it is (using 'tr' at
db>)

> 
> Just for a test I am gonna disable SMP and HT too. (I
have set them back
> since others suggested that its ok for dom0 and my
problem was the memory
> anyway) Also I will try linux domU but that is not my
goal so gotta get
> BSD going 
> 
> 
> > It would give a hint about what the dom0 or domU
is doing, though. A top
> > on your desktop should show what process is
becoming crasy, isn't it ?
> 
> Okay, will investigate that as well.
> 
> > I never played with this either; but I think it's
with
> > xm sched-sedf <domain> -s
> > you can try
> > xm sched-sedf <domain> -s 10
> 
> Ok I will try this too but rather fix the root cause...
If it's not needed
> for most cases I wouldn't be happy to use it. It would
be perhaps just a
> bypass...

If it's the domU which is looping, this won't help.

-- 
Manuel Bouyer <bouyerantioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la
difference
--

On Fri, Sep 15, 2006 at 10:39:19PM +1200, variianiwa.co.nz wrote:
> Sorry for the confusions...
> 
> > Not sure what you mean here
> http://w
iki.xensource.com/xenwiki/NetBSDdomU
> 
> Before the first reboot there's a suggestion to:
> mount /dev/xbd0a /mnt
> cd /mnt/dev
> cp -pR /dev/rxbd* .
> cp -pR /dev/xbd* .
> halt -p
> 
> It was about an older NetBSD 2 domU install but anyway
I have actually
> tried without this however same hanging occur.

It's not needed any more

> 
> > I don't understant what you mean here. Is the Xen
box you desktop too ?
> > Otherwise how could a ssh session make your
desktop swap ?
> > When you says the CPU is 100%, is it dom0's CPU
or domU's CPU ?
> 
> No. I log into the dom0 via ssh (server room is a bit
away+cold+noisy...)
> >From the ssh console running on the dom0 I start up
the VM as root with -c
> paramter of course then I log into the VM as root. 5
minutes later it
> stops responding. In the meantime I have another dom0
ssh connection and
> that shows the CPU load to be 100% and the fully used
512M memory by xm
> top.

Who is at 100%, dom0 or the domU ? The fast that the 512M
are all used
may not be a problem.

> 
> When this hanging happens it drives my desktop Fedora
nuts. It becomes non
> responsive too and start swapping... I didn't
understand this neither but
> let's not worry about my dekstop. 

It would give a hint about what the dom0 or domU is doing,
though. A top
on your desktop should show what process is becoming crasy,
isn't it ?

> 
> > Xen's scheduler parameters. Changing this to have
a non-0 minimum time
> > slice
>  Could you point me to the right place about it? I
don't understand this.

I never played with this either; but I think it's with
xm sched-sedf <domain> -s 
you can try
xm sched-sedf <domain> -s 10

-- 
Manuel Bouyer <bouyerantioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la
difference
--