NetView's netmon method of limiting queries to router's routing table is
done on a per router bases. If you look at the xnmsnmpconf (SNMP
Configuration) panel, the field "Number of Route Entries" defines the
number of routing table entries to query before terminating the walk
through the routing table. By default I believe this value is 300.
You must also remember why NetView is walking the routing table, to
discover how you have defined the IP routing within your network, and
thereby determining if there are any additional routers in your network to
be discovered.
So setting the value on initial discovery to a small value may cause parts
of the network not be automatically discovered. But after you have
discovered the network and are a sure you have the entire network you wish
to manage, there is not reason why you can not decrease this value. I
would NOT set it to zero but 100 should be small enough to prevent from
over-loading the router.
The question I have for router vendors, is why is a snmpgetnext query
causing such a high CPU usage, is it a problem with the IOS of the router?
>Scott is right about netmon for HP Openview. The same problem doesn't
apply
>to IBM Netview, where netmon uses a different strategy on polling.
>
>Kostas Kottos,
>ASYK
>From: Scott Wilson <swilson@RPM.COM>
>Date: P]lptg, 3 Dejelbq_ou 1998 8:50 ll
>Subject: Re: Netmon/Netview Pounding of Routers
>With OpenView it is a netmon option (I believe -r) to stop checking the
routing
>tables on routers. This is usually what puts the hit on the routers CPU.I
do
>not think that the same option is available on NetView however (I wish it
>were). What people have suggested here before is to block that section of
the
>MIB on the device (this is of course device dependent) so that your
NetView
>system is not configured with a valid comm string for the route able
portion of
>the devices MIB, while still being able to see the rest of the MIB.
>
>Connie Logg wrote:
>
>> I have noticed for some time that one of our routers will have sustained
cpu
>> utilization of greater than 90% for many hours (sometimes a day or more
at a
>> spell). I have suspected netmon/netview, and recently unmanaged the
>> routers. It seemed to resolve the problem.
>>
>> The November 23 issue of Network World discusses the problem (HP
Openview),
>> and indicates that there is a way around the problem without unmanaging
the
>> routers. There were no details although for netview.
>>
>> Can someone please tell me how to resolve this problem without leaving
the
>> routers unmanaged?
>>
|