On Thu, Dec 15, 2005 at 09:32:51AM +0100, Xavier Montagutelli wrote:
> On Thursday 15 December 2005 05:33, Matt Gorecki wrote:
> > Xavier Montagutelli wrote:
> > >Hello list,
> > >
> > >I have added a HowTo on the Wiki (Documentation -> HowTo section) to show
> > > a way to graph vserver-stat output with cacti.
> > >
> > >http://linux-vserver.org/Cacti+configuration
> > >
> > >It's quite long, but I hope it's still useable. It has been a bit hard to
> > > make this setup, so I think it can be of interest.
> > >
> > >Feel free to correct. Or even remove if you think it's useless... You can
> > > also contact me if you have problems with the setup (I guess these
> > > questions would be off-topic on this list ?), even if cacti remains
> > > "black magick" for me after this trial ...
> > >
> > >HTH,
> >
> > Very nice, this will be very useful.
> >
> > I have been wondering about the best way to monitor network traffic for
> > each individual vserver. Have you done something similar for those
> > statistics as well?
>
> No, because in my case memory is the most demanded resource.
>
> I use the standard SNMP data (already defined in cacti for example)
> for network traffic => I get data per *physical* interface. I
> don't know how to get network traffic per IP address by SNMP.
> Perhaps the only way to do that is by exploiting data from
> /proc/virtual/<ctx>/cacct and using a similar trick as I did (through
> some shell script).
> (http://linux-vserver.org/HowTo+Read+ProcFS)
if you want _network_ data per context, your best choice
is probably an iptables accounting rule based on the guest
IP(s)
> The only difference in the how to would be on the "Getting
> vserver-stat output" section :-) If you find it difficult and you are
> really interested, I can spend some time on this.
>
> I also use the standard SNMP data for load average on the host, not
> by vserver (we could perhaps get this data using the "virt_load" flag
> on all vservers and a shell script on the host). But this information
> doesn't seem very relevant, just for cosmetic. Or would there be a
> real interest ?
the /proc/virtual/<xid>/* contains a bunch of useful
information, but for an efficient monitoring system some
faster communication methods should be considered ...
let me know if there is real demand, and somebody is
willing to do the userspace part .. we could then go for
netlink or relayfs ...
best,
Herbert
> > Matt
> > _______________________________________________
> > Vserver mailing list
> > Vserver@list.linux-vserver.org
> > http://list.linux-vserver.org/mailman/listinfo/vserver
>
> --
> Xavier Montagutelli
> Service Commun Informatique
> Universite de Limoges
> Tel : +33 555457720
> Cle GPG : http://pgp.mit.edu 1024D/175CE198
> _______________________________________________
> Vserver mailing list
> Vserver@list.linux-vserver.org
> http://list.linux-vserver.org/mailman/listinfo/vserver
_______________________________________________
Vserver mailing list
Vserver@list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver
Received on Thu Dec 15 14:45:57 2005