Hola! :-)
Thanks to all for your answers, I've included them here below. However, I
ended up implementing a fix on the router after discussions with our CCIE on
staff. Our "fix" is to configure the router to ignore requests for the MIB
object by setting up SNMP "views". I added the following commands to our
router:
snmp-server view <view-name-here> mib2 included
snmp-server view <view-name-here> private included
snmp-server view <view-name-here> ip.21 excluded
snmp-server community <community-string> view <view-name-here> RO
[access-list]
snmp-server community <community-string> RW [access-list]
The first two lines allow mib2 queries and cisco mib queries (from the
private branch). The 3rd line causes the router to ignore queries for
ipRouteTable (*very* large when connected to the Internet via BGP ;). The
last two lines allow read-only and read-write SNMP queries. I didn't use a
view in the last line as an example to show you can mix and match.
Stephen Houle suggested:
>
>Go to the menu pick "Options -> SNMP Configuration". Enter your routers in
>the "Specific Nodes" list (use the ip address or name matches their
>selection name). Then for each record, set "Configuration Polling
Interval"
>to something larger than the 1 day that is the default. Also turn off
>"Discovery Poll".
>
>You can also turn off "Auto Adjust", but leave "Discovery Poll" on, and
then
>set a large "Fixed Polling Interval" (which should say "Fixed Discovery
>Polling Interval", and set it to 1 day or something). This will allow you
>to still do discovery polling (if you need it), but not at the 15 minutes
>(or whatever) that Netview does it by default and/or on an auto-adjusted
>schedule.
>
>I don't think you can prevent a demand poll from happening if a user does
>one, but this will allow you to slow thing down a little bit.
And Tivoli folks said:
>
>specifying (via Options->SNMP) a bogus
>community string would prevent the node in question
>from being hammered with SNMP requests; you can even
>configure the SNMP retry count to be 1 so that netmon
>won't do excessive retries using the bogus password.
>
>Other than this, all you can do is to UNmanage the
>node, but I don't believe that this is what you
>are looking for.
>
>The bogus password/low retry count will be used by
>netmon for the configuration polls, new-node polls, and
>any demand polls instantiated by the user (via
>Test->Demand Poll).
>
And the original question was:
>
> Hola! :-)
>
> Is there a way to configure netmon to "skip" demand polls and/or route
> table queries for specific nodes?
>
> After installing NetView NT I noticed that CPU utilization on two routers
> exceeded 90% because of the SNMP process. After doing some research, I
> noticed it was whenever netmon did a demand poll, or I used MIB browser to
> query for route table information, that caused the high cpu utilization.
>
> I've thought of changing the object attributes to be something other than
> router, but I was hoping there might be another method. Also, I don't
> believe it's a router IOS bug (Cisco 7513 running 11.2(13)P) because the
> routers in question connect to the Internet (big I) and are running BGP-
> i.e., large route table = large SNMP process when queried.
>
> Thanks in advance for the help!
>
> Stephen Truxton
> ______________________________
> Intel Corporation
> Hillsboro, Oregon
> Email: stephen.truxton@intel.com
> I'm *NOT* an attorney, so I'm *NOT*
> speaking for Intel...
>
>
|