08-19-2004 11:20 PM - edited 03-02-2019 05:53 PM
hello everbody,
i'm still looking for a solution for my problem.
i've got to do something like proactive management...
i want to get informed about any problem on a router or it's interfaces before it is noticed by some users on the network.
you know, there are some things that happen before the performance gets less or the interface goes in error disable,
ok, let's step forwards and come to my question..
what are these counters or stats or whatever that may tell me that anything bad will happen soon ?
how can i get informed, snmp, traps, rmon ?
kind regards
chris
08-20-2004 12:58 AM
We make extensive use of Big Brother (see http://www.bb4.org/ ). It monitors all sorts of systems including Cisco. In conjunction with Routermon (see http://www.routermon.net/ ), you can monitor routers and switches and alert at different threshold levels.
Here is a very small example of the output :-
yellow Fri Aug 20 09:51:56 CET 2004 [d1-b31l-2]
: System uptime: 22 days, 2:24:34 (hh:mm:ss)
: CPU usage is high: 50% (thresholds: yellow=50%:red=90%)
System description:
-------------------
Cisco Internetwork Operating System Software
IOS (tm) C3500XL Software (C3500XL-C3H2S-M), Version 12.0(5)WC3b, RELEASE SOFTWARE (fc1)
Copyright (c) 1986-2002 by cisco Systems, Inc.
Compiled Fri 15-Feb-02 10:51 by antonino
etc
This is for a switch that is reporting occasional high CPU. It also stores a history all devices as well, so you can see when a problem started occurring.
Hope that is of some help.
Pete
08-20-2004 01:08 AM
thanks for your message pete,
i'm sorry but it realy doesn't help...
in fact i'm not looking for some management tools, but for any best practice advisory for example what counters should not leave zero status or reach any value in a specified time...
hope my request is much more understandable now
kind regards
chris
08-20-2004 01:48 AM
Yes - I see what you are after now.
The real problem with this is that there aren't really any hard-and-fast rules about a lot of the counters. In we take a couple of examples :-
1. On any interface, the CRC error counter should always be zero, but then if you get a few CRC errors, do you spend time investigating something which is a relatively minor problem? Also, what figure do you take as being the threshold and when does it become a more serious problem?
2. The collision counter on a half-duplex ethernet interface is likely to be non-zero, especially if the utilisation is high, because collisions are part of the normal operation of ethernet. But again, what to you say is an acceptable value, and where do you set your thresholds?
This is a very difficult area to pin down I reckon, because the whole question of thresholds is fairly subjective. I've just had a quick scan on Google, but most of the replies are fairly high-level and don't get down to the sort of level that you are after. They generally talk about this like how to set up Network Management, SLAs and so on.
So, sorry not to be much help here and good luck in your quest.
Pete
08-20-2004 02:58 AM
yeah, you got point
that's what i'm talkin' about.
think it'll be my task to look for network problems and what happend before on the interfaces.
thanks pete
kind regards
chris
08-20-2004 06:05 AM
One counter that should never be above zero would be queue drops for a given hardware interface on a router. The snmp variabls are locIfInputQueueDrops (1.3.6.1.4.1.9.2.2.1.1.26) and locIfOutputQueueDrops (1.3.6.1.4.1.9.2.2.1.1.27). These apply to routers running IOS. If these are above zero, you might have a buffer leak. Once you know how to set up mrtg, you can monitor and graph these specific snmp variables.
08-22-2004 11:14 PM
thanks john,
OIDs are always good to know, i'll try some polling later that day in the lab.
kind regards
chris
08-23-2004 06:25 AM
There is some useful stuff about counters and network monitoring in the back of the CCIE Network Design and Case Studies book, Chapter 14:
http://www.amazon.com/gp/reader/1578701678/ref=sib_dp_pt/102-5068235-6748957#reader-link
Unfortunately this book seems to be out of print now, but I reckon it is well worthwhile trying to get hold of a copy.
Kevin Dorrell,
Luxembourg
08-24-2004 02:30 AM
i've found something else to tell me the oid of a specified output.
http://www.cisco.com/cgi-bin/Support/Mibbrowser/unity.pl
the only things missing are thresholds for these outputs.
kind regard
chris kulinski
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide