[mrtg] Re: provisioning question for a new MRTG server

[Barber, Erik A.]

Try setting off 20% of the cron jobs in 1 minute increments, ie 20% at 1:00,
20% at 1:01, etc; that should help spread out the load a bit.  I'm
monitoring about 1100 target this way on an Ultra 10, and the load only gets
up to about 3.  Not bad for 4K.

-----Original Message-----
From: Jon Heugh [mailto:jon at heugh.com]
Sent: Thursday, June 06, 2002 12:11 PM
To: Greg.Volk at edwardjones.com
Cc: mrtg at list.ee.ethz.ch
Subject: [mrtg] Re: provisioning question for a new MRTG server



I have MRTG on a SUN 420R with 2 processors and 4GB of RAM.
I monitor around 420 routers for a total of 4826 Interfaces.
I spawn a MRTG process for each router.
When I had these cron'ed for 5 min the box would spike to a load of 20 and
would be unreliable.
I was forced to cron them at 10 min intervals.  Now the load will spike to
about 8 at times.
I was expecting more out of our SUN box, but hey, nothing that another 50K
cant cure.


----- Original Message -----
From: <Greg.Volk at edwardjones.com>
Cc: <mrtg at list.ee.ethz.ch>
Sent: Wednesday, June 05, 2002 6:06 PM
Subject: [mrtg] Re: provisioning question for a new MRTG server


>
> >
> > Greg
> >
> > This is the type of info I was looking for.  Right now on
> > my old PIII 800MHz system it takes ~ 12 minutes to poll
> > ~10,000 ports.  I also keep the cfgs broken up by network
> > device.
> >
>
> I'm glad I could help.
>
> > BTW, that was 10,000 active ports, off hand if I included
> > all our ports, whether up or down, the number would probably
> > double.  This may actually occur because there is conversation
> > about turning on all ports on the Cat's, putting the unused
> > one in an unrouted LAN.  This would allow someone to move into
> > an office, connect system to wall jack, using DHCP they'd
> > end up with an IP in the private LAN, goto registration
> > website to get routable IP.  Once registered the VLAN info
> > would be updated on the switch within X minutes and voila
> > they'd be on the net.
> >
>
> I would really like to hear from someone who is doing, or at
> least attempted to do 10,000+ five minute targets on a recent,
> multi-proc x86 platform. For as much as that 220R costs, I
> expect it to do everything and make me breakfast. But
> seriously, I could have bought three really sweet x86 systems
> with all kinds of bells and whistles, for the same money, and
> my dynamic graph generation would be a lot faster. Of course
> if the x86 doesn't have the requisite IO capability to deal
> with 18,000 targets every five minutes, then graph generation
> is a moot point.
>
>
> --
> Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
> Archive     http://www.ee.ethz.ch/~slist/mrtg
> FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org
> WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi
>


--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi

--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi