When host Netview shows instantaneous status of SNA connections to nodes,
it is because the host is holding one end of each session, and knows
immediately when the other end is dropped (well, that's how I think of
it, anyway). With the IP product's polling architecture, it is a whole
other
thing. If I were doing it, I would look for an agent to run on each node
that monitored the apps every couple of seconds, and only reported
changes to Netview. Netview should of course keep in touch with
that agent to make sure it is really up, but that can be much less
frequent.
What I have just described sounds a lot like the MLM (or SIA) on each
target
node. As soon as the app fails (or however many seconds you set the
local checking for), a trap goes to Netview and status is changed.
Just getting IP status on a realtime basis would be iffy. Doing anything
more than that I would expect problems.
Cordially,
Leslie A. Clark
IBM Global Services - Systems Mgmt & Networking
Detroit
====================================================
So you are saying that all the things you monitor are local to you, and
that is
why a 5-second cycle might work? I'd still be starting with a 1 minute
cycle,
or 30 seconds or even 15 seconds, before I went to 5 seconds. This is
still a
network we are talking about and not the same box. And if your FDDI is
not
local then you will probably want a cycle of minutes and not seconds.
But
have at it. Experiment all you like. Watch your CPU cycles.
You change the NetView polling cycle by using xnmsnmpconf, either by
typing
that in on the command line, or by using the GUI and pulling down Options
-->
SNMP Configuration. Is that what you did? It takes netmon awhile to
notice
that the cycle has been shortened (because he has already established
queues and
timers for the old values, and won't build new ones until the old ones
expire).
So after you dramatically lower the values you may want to ovstop netmon
and
then ovstart him.
Good luck
James Shanks
Tivoli (NetView for UNIX) L3 Support
Bob Stamm <Robert_Stamm@RES.RAYTHEON.COM> on 02/23/2000 09:23:48 AM
Please respond to Discussion of IBM NetView and POLYCENTER Manager on
NetView
<NV-L@UCSBVM.UCSB.EDU>
To: NV-L@UCSBVM.UCSB.EDU
cc: (bcc: James Shanks/Tivoli Systems)
Subject: Re: Demand Polling
We at Raytheon build Air Traffic Control Systems. My attempt to use
NetView is in this context where
we have a system with perhaps 60 sun workstations driving large 2kx2k
displays presenting aircraft into
to controllers. Along with these are a series of about 20 servers that
supply a redundant FDDI network with
flight info drom a database, radar data, and a bunch of other stuff.
Our legacy "Command and Monitoring Display" is a Raytheon-built application
that speaks with "agents"
that run on every box to derive the status of all these system components.
This status is displayed using
an X-based application that has a set of icons whose colors show the status
of the applications running on
each of these network boxes. At the present time it can display a change
in the status of any of the
applications running in this system within 3 seconds.
Running on the same workstation as the CMD I've installed NetView. I've
done what I can to midify NetViews
GUI to look like our legacy CMD and all that. I use NetView to view the
status of the hardware on which each
of the system's applications runs. If I was to shut down a node the legacy
CMD would show the appropriate
icons turning red within 3 seconds. By the time NetView shows anything the
technician has enough time to
walk across the room to the box to see what's wrong! NetView may tale 5
minutes to show there;s a problem.
For a critical system like Air Traffic Control this is unacceptable.
So therefore I want to diminish the time between pings to make NetView's
performance appear to be reasonable.
But I'm not sure how to get this to happen successfully.
As far as how quickly the boxes will resond to the ping, I'm not exactly
sure but when pinged manually they repond
immediately to the ping.
So in summary the network is not hugh but it's not very small either.
What, if anything can I do?
As a matter of fact Tivoli has given us a license to try my work out early
next week in Munich Germany as
part of a larger overall demonstration to a potential customer in Europe.
So this is sortof important to us.
James Shanks <James_Shanks@TIVOLI.COM> on 02/22/2000 10:34:27 PM
Please respond to Discussion of IBM NetView and POLYCENTER Manager on
NetView <NV-L@UCSBVM.UCSB.EDU>
To: NV-L@UCSBVM.UCSB.EDU
cc: (bcc: Robert Stamm/RES/Raytheon/US)
Subject: Re: Demand Polling
5 seconds for a default? Bob, that's going to drive your network and your
NetView box to its knees. Think about that. A ping to every device in
your
database every five seconds? How quickly will they respond? Before long
netmon
will probably behind in one cycle when it becomes time to start another,
resulting in 100% CPU utilization and a network congested with ICMP
requests.
Unless you have a very tiny network this is unworkable.
So let's back up. What exactly are you trying to do? How did you try to
change
the polling interval? And what exactly didn't work?
James Shanks
Tivoli (NetView for UNIX) L3 Support
Bob Stamm <Robert_Stamm@RES.RAYTHEON.COM> on 02/22/2000 05:30:59 PM
Please respond to Discussion of IBM NetView and POLYCENTER Manager on
NetView
<NV-L@UCSBVM.UCSB.EDU>
To: NV-L@UCSBVM.UCSB.EDU
cc: (bcc: James Shanks/Tivoli Systems)
Subject: Demand Polling
By default Netview polls at a rate of once every 5 minutes. I've attempted
to change this (I'd like
it to ne once per 5 seconds) but I can;t get ot to work. Does anyone know
the trick?
Bob
Raytheon Company
|