Leslie,
I'm so confused! I didn't realize it but I added a -K 2 (capital K) to the
netmon.lrf file. It was previously at a -K 0 (capital K). My users did not
like the white unreachable routers or network objects. They also wanted to
see all events in the event viewer.
If I understand the RouterFaultIsolation.htm document correctly, the -K 1 or
-K 2 (capital K) enable RFI and then allow you to use -k 2 (little k)
options. The -k 2 (little k) is not an independent option. Correct? And I
cannot use -K 0 (capital K) in conjunction with -k 2 (little k). Correct?
Here's what happened and why I'm asking about reduced polling. We have
noticed that our Cisco 3640 frame relay router was being clobbered by SNMP
requests. The 3640 has 3 "AD" interfaces. The 7206 has 2 "AD" interfaces.
Two remote routers on the 3640 frame circuits have "AD" interfaces. The 7206
CPU had no problems with all these SNMP requests. On the other hand, the
3640 CPU could not cope.
On Sunday I did a trace of the Cisco 3640 and a trace of a Cisco 7206
routers. Traces showed an echo request from NV, an echo reply, another echo
request from NV to an "AD" interface then the router sent a network
unreachable. The router then sent a flood of SNMP GetResponses from the
router. The unreachable occurred on a 5 minute cycle (default status poll
interval) and then the SNMP GetResponses flooded in from the router. When
the router completed the SNMP GetResponses for each of the "AD" interfaces
it was just about time for the 5 minute status poll of the 1st "AD"
interface to occur again. As soon as I set -K 2 in netmon.lrf - the pattern
stopped. The pings and subsequent network unreachables occur on the 5 minute
cycle but are not followed by a flood of SNMP GetResponses.
What do you think?
Thanks very much for your response.
Ray Westphal.
Enterprise Rent-A-Car.
> Well, the way I read /usr/OV/doc/RouterFaultIsolation.htm, the recovery is
> started by a successful status poll of the root-cause router. Setting -k 2
> (that's the little k, not the big K) to reduce polling to unreachable
> routers
> does not affect the polling of the router that is only Marginal, the one
> that
> is causing the perceived outage. The affected area would remain
> white until the regular 5-minute (or whatever) poll of the down interface
> on that device gave netmon cause to initiate the recovery process, and
> go try everything again. Or you can ping something, and if it works, the
> recovery starts.
> That said, I don't understand why netmon would be doing continuous
> polling of the devices with the admin-down interfaces. Although I do
> understand the intuitive action you took. Are you sure that your
> netmon settings are now -K 1, -k 2? Or are you using only -K 2, which
> depending on how the code is written might just mean that you
> turned off RFI altogether (eg. -K 1 is on, anything else is off)
> Cordially,
Leslie A. Clark
IBM Global Services - Systems Mgmt & Networking
|