VM id in KVM?

VM id in KVM?

[Jun Koi]

Hi,

Do you think that it is necessary to have something like id for a
specific VM on KVM? Without it, how can we identify a VM? By qemu's
pid? Is that good, as we then depend on a particular device model?

Thanks,
Jun

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Avi Kivity]

Jun Koi wrote:
> Hi,
>
> Do you think that it is necessary to have something like id for a
> specific VM on KVM? Without it, how can we identify a VM? By qemu's
> pid? Is that good, as we then depend on a particular device model?
>   

 From a Linux point of view, the pid identifies the VM.  A management 
application can, however, use its own VM identifiers as it sees fit, and 
map the (possibly persistent, gloablly unique, and ridiculously long) 
VMID to the pid.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Carsten Otte]

Avi Kivity wrote:
>  From a Linux point of view, the pid identifies the VM.  A management 
> application can, however, use its own VM identifiers as it sees fit, and 
> map the (possibly persistent, gloablly unique, and ridiculously long) 
> VMID to the pid.
It might be preferable to have something that is persistent over guest 
migration. Makes life easier for the management application as far as 
I see.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Avi Kivity]

Carsten Otte wrote:
> Avi Kivity wrote:
>>  From a Linux point of view, the pid identifies the VM.  A management 
>> application can, however, use its own VM identifiers as it sees fit, 
>> and map the (possibly persistent, gloablly unique, and ridiculously 
>> long) VMID to the pid.
> It might be preferable to have something that is persistent over guest 
> migration. Makes life easier for the management application as far as 
> I see.

It may make sense to add a vmid to qemu (or to keep it in the management 
application entirely).  Certainly the kernel doesn't need to know about it.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Avi Kivity]

Avi Kivity wrote:
> Carsten Otte wrote:
>> Avi Kivity wrote:
>>>  From a Linux point of view, the pid identifies the VM.  A 
>>> management application can, however, use its own VM identifiers as 
>>> it sees fit, and map the (possibly persistent, gloablly unique, and 
>>> ridiculously long) VMID to the pid.
>> It might be preferable to have something that is persistent over 
>> guest migration. Makes life easier for the management application as 
>> far as I see.
>
> It may make sense to add a vmid to qemu (or to keep it in the 
> management application entirely).  Certainly the kernel doesn't need 
> to know about it.
>

I take it back.  The only entity that can enforce uniqueness is the 
management application, therefore that should be the entity that knows 
about them.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Carsten Otte]

Avi Kivity wrote:
> It may make sense to add a vmid to qemu (or to keep it in the management 
> application entirely).  Certainly the kernel doesn't need to know about it.
Yes, I agree that should be userland only.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Daniel P. Berrange]

On Mon, Jul 23, 2007 at 02:19:41PM +0300, Avi Kivity wrote:
> Avi Kivity wrote:
> > Carsten Otte wrote:
> >> Avi Kivity wrote:
> >>>  From a Linux point of view, the pid identifies the VM.  A 
> >>> management application can, however, use its own VM identifiers as 
> >>> it sees fit, and map the (possibly persistent, gloablly unique, and 
> >>> ridiculously long) VMID to the pid.
> >> It might be preferable to have something that is persistent over 
> >> guest migration. Makes life easier for the management application as 
> >> far as I see.
> >
> > It may make sense to add a vmid to qemu (or to keep it in the 
> > management application entirely).  Certainly the kernel doesn't need 
> > to know about it.
> >
> 
> I take it back.  The only entity that can enforce uniqueness is the 
> management application, therefore that should be the entity that knows 
> about them.

When managing QEMU & KVM guests, libvirt provides 3 identifiers with
varying levels of uniqueness

 - ID - a integer unique amongst all active guests on a host
 - Name - a string uninque amongst all active & inactive guests on a host
 - UUID - 32 byte hex string unique globally

We don't expose the PID directly of the QEMU binary directly. Name and UUID
are both stable across migration - the ID changes upon migration. As Avi 
says I dont't see how a individual QEMU process could provide any meaningful
identifier itself aside from its PID whose uniqueness is guarenteed by the
OS on its behalf.

Regards,
Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Avi Kivity]

Daniel P. Berrange wrote:
> When managing QEMU & KVM guests, libvirt provides 3 identifiers with
> varying levels of uniqueness
>
>  - ID - a integer unique amongst all active guests on a host
>  - Name - a string uninque amongst all active & inactive guests on a host
>  - UUID - 32 byte hex string unique globally
>
>   

Does libvirt also provide a lighter-weight interface that doesn't know 
about names and uuids?  I imagine a cluster-wide management solutions 
will want to keep all configuration in a central database and just tell 
libvirt "start a guest with this configuration".


-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Daniel P. Berrange]

On Mon, Jul 23, 2007 at 04:39:33PM +0300, Avi Kivity wrote:
> Daniel P. Berrange wrote:
> >When managing QEMU & KVM guests, libvirt provides 3 identifiers with
> >varying levels of uniqueness
> >
> > - ID - a integer unique amongst all active guests on a host
> > - Name - a string uninque amongst all active & inactive guests on a host
> > - UUID - 32 byte hex string unique globally
> 
> Does libvirt also provide a lighter-weight interface that doesn't know 
> about names and uuids?  I imagine a cluster-wide management solutions 
> will want to keep all configuration in a central database and just tell 
> libvirt "start a guest with this configuration".

Name is mandatory, if you omit the UUID it will generate one for its own
internal tracking purposes. You can also create a guest without having to
pre-define a config file - it will disappear with no trace once the guest
is shutdown.

Regards,
Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Rusty Russell]

On Mon, 2007-07-23 at 14:32 +0100, Daniel P. Berrange wrote:
> As Avi 
> says I dont't see how a individual QEMU process could provide any meaningful
> identifier itself aside from its PID whose uniqueness is guarenteed by the
> OS on its behalf.

Indeed.  I use it in my pr_guest() patch which replaces all those
printks with:

+/* The guest did something wrong/strange. */
+#define pr_guest(vcpu, fmt, ...)					\
+	do {								\
+		if (__printk_ratelimit(5*HZ, 10))			\
+			printk(KERN_WARNING "kvm: %i: cpu%i " fmt,	\
+			       (vcpu)->kvm->pid, (vcpu)->vcpu_id	\
+			       , ## __VA_ARGS__);			\
+	} while(0)

I'll polish it up and send it on.  I don't want Avi getting bored.

Cheers,
Rusty.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Jun Koi]

On 7/24/07, Rusty Russell <rusty-8n+1lVoiYb80n/F98K4Iww@public.gmane.org> wrote:
> On Mon, 2007-07-23 at 14:32 +0100, Daniel P. Berrange wrote:
> > As Avi
> > says I dont't see how a individual QEMU process could provide any meaningful
> > identifier itself aside from its PID whose uniqueness is guarenteed by the
> > OS on its behalf.
>
> Indeed.  I use it in my pr_guest() patch which replaces all those
> printks with:
>
> +/* The guest did something wrong/strange. */
> +#define pr_guest(vcpu, fmt, ...)                                       \
> +       do {                                                            \
> +               if (__printk_ratelimit(5*HZ, 10))                       \
> +                       printk(KERN_WARNING "kvm: %i: cpu%i " fmt,      \
> +                              (vcpu)->kvm->pid, (vcpu)->vcpu_id        \
> +                              , ## __VA_ARGS__);                       \
> +       } while(0)
>
> I'll polish it up and send it on.  I don't want Avi getting bored.
>

Looks like you will add pid into kvm struct? That is cool, and really
what I am looking for :-)


Thanks,
Jun

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Avi Kivity]

Rusty Russell wrote:
> On Mon, 2007-07-23 at 14:32 +0100, Daniel P. Berrange wrote:
>   
>> As Avi 
>> says I dont't see how a individual QEMU process could provide any meaningful
>> identifier itself aside from its PID whose uniqueness is guarenteed by the
>> OS on its behalf.
>>     
>
> Indeed.  I use it in my pr_guest() patch which replaces all those
> printks with:
>
> +/* The guest did something wrong/strange. */
> +#define pr_guest(vcpu, fmt, ...)					\
> +	do {								\
> +		if (__printk_ratelimit(5*HZ, 10))			\
> +			printk(KERN_WARNING "kvm: %i: cpu%i " fmt,	\
> +			       (vcpu)->kvm->pid, (vcpu)->vcpu_id	\
> +			       , ## __VA_ARGS__);			\
> +	} while(0)
>
> I'll polish it up and send it on.  I don't want Avi getting bored.
>
>   

With the ioctl interface, vms aren't really tied to tasks.

How about current->pid (or with struct pid and containers, probably
current->something->something->pid)?

-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/

Re: VM id in KVM?

[Rusty Russell]

On Tue, 2007-07-24 at 08:17 +0300, Avi Kivity wrote:
> With the ioctl interface, vms aren't really tied to tasks.

True, but this is really a debugging aid.

> How about current->pid (or with struct pid and containers, probably
> current->something->something->pid)?

I was reluctant because there's nothing inherent in this function which
says it's the current CPU's guest we're complaining about.

...Although, it always is.  OK, I'm convinced.

Cheers,
Rusty.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/