3:32 p.m.
After playing around with Avahi, I have now a server that is able to query
all hosts on the network and their VMs, I think it would be interesting to
publish utilisation statistics in the Txt record of the service.
Of course one could login to a host and fetch the info, but I think it
would be (managementwise) interesting if the hosts would announce this
information.
The implication for libvirtd would be a thread that updates
lets say, per minute the avarage cpu load / free memory.
I might even go further and publish all resource statistics over mDNS.
Maybe people have ideas about this too, securitywise i'm not really
interested in the implications since I'm publishing it on a management
LAN.
Stefan
3:42 p.m.
On Wed, May 14, 2008 at 10:32:45PM +0200, Stefan de Konink wrote:
After playing around with Avahi, I have now a server that is able to
query
all hosts on the network and their VMs, I think it would be interesting to
publish utilisation statistics in the Txt record of the service.
[snip]
I might even go further and publish all resource statistics over
mDNS.
Maybe people have ideas about this too, securitywise i'm not really
interested in the implications since I'm publishing it on a management
LAN.
This is already a solved problem - any resource monitoring system can
publish this kind of information. Rich Jones has already written libvirt
plugins for both collectd and nagios to publish VM information, and so
there's no need to re-invent this in libvirtd itself, nor abuse Avahi/mDNS
service discovery for this purpose
http://collectd.org/documentation/manpages/collectd.conf.5.shtml#plugin_l...
http://et.redhat.com/~rjones/nagios-virt/
Regards,
Daniel.
--
|: Red Hat, Engineering, Boston -o- http://people.redhat.com/berrange/ :|
|: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
4:13 p.m.
On Wed, 14 May 2008, Daniel P. Berrange wrote:
> I might even go further and publish all resource statistics over
mDNS.
> Maybe people have ideas about this too, securitywise i'm not really
> interested in the implications since I'm publishing it on a management
> LAN.
This is already a solved problem - any resource monitoring system can
publish this kind of information. Rich Jones has already written libvirt
plugins for both collectd and nagios to publish VM information, and so
there's no need to re-invent this in libvirtd itself, nor abuse Avahi/mDNS
service discovery for this purpose
http://collectd.org/documentation/manpages/collectd.conf.5.shtml#plugin_l...
http://et.redhat.com/~rjones/nagios-virt/
I'm currently implementing resource balancing between hosts. So one way or
the other I want to receive or pull from the active hosts in the network
their active states. And migrate host from heavy used machines to unused
ones. Also, bias new started VMs to an unused machine.
I see the collectd plugin takes the opposite approach which need to
specify information while this is unknown for my statistics management.
What I made is currently an avahi-based-libvirt-statistics-fetcher. And I
presume the best situtation would be if it was an
avahi-based-libvirt-statistics-receiver.
If I'm reinventing the wheel because ovirt does this already or something
like it... please tell me.
Stefan
12:21 p.m.
On Wed, May 14, 2008 at 11:13:40PM +0200, Stefan de Konink wrote:
On Wed, 14 May 2008, Daniel P. Berrange wrote:
> > I might even go further and publish all resource statistics over mDNS.
> > Maybe people have ideas about this too, securitywise i'm not really
> > interested in the implications since I'm publishing it on a management
> > LAN.
>
> This is already a solved problem - any resource monitoring system can
> publish this kind of information. Rich Jones has already written libvirt
> plugins for both collectd and nagios to publish VM information, and so
> there's no need to re-invent this in libvirtd itself, nor abuse Avahi/mDNS
> service discovery for this purpose
>
> http://collectd.org/documentation/manpages/collectd.conf.5.shtml#plugin_l...
> http://et.redhat.com/~rjones/nagios-virt/
I'm currently implementing resource balancing between hosts. So one way or
the other I want to receive or pull from the active hosts in the network
their active states. And migrate host from heavy used machines to unused
ones. Also, bias new started VMs to an unused machine.
I see the collectd plugin takes the opposite approach which need to
specify information while this is unknown for my statistics management.
I don't know what you mean here ? Collectd will push out information for
all guests. It can be configured to sned it in a variety of ways, and if
none of the existing data transports are suitable it has extensive plugin
capabilities. mDNS is also a push protocol, so there's really no difference
between what you could do with mDNS vs collectd at your receiving end.
If I'm reinventing the wheel because ovirt does this already or
something
like it... please tell me.
Well oVirt is a closed-loop system - it has a standardized host OS image
for nodes running VMs, and a matched admin host to receive - oVirt isn't
intended to provide generic services for everyone wanting to us libvirt.
Its just one of many apps built ontop of libvirt. As it hapens oVirt
currently uses collectd to send resource & guest utilization data to its
management server. We may or may not keep doing this, or change to a
diffferent resource collection daemon that uses the AMPQ message bus..
Dan.
--
|: Red Hat, Engineering, Boston -o- http://people.redhat.com/berrange/ :|
|: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
6035
days inactive
6036
days old
3 comments
2 participants
participants (2)
-
Daniel P. Berrange -
Stefan de Konink