RE: [nv-l] Polling Question

[James Shanks <jshanks@us.ibm.com>]

netmon configuration is not my area of expertise, but I can offer this.

> I need to poll some devices more frequently than every 20 minutes.
You can configure the polling intervals using Option--> SNMP from the
NetView GUI or by typing snmpconf from a command line.
You can do that for specific nodes, or change the global default.  Those
nodes which do not override will use the global default.
I would have thought you had been using that already, given the number of
objects you have.

You can set  a polling frequency (Poll every x units)  and a time-out value
(Retry SNMP requests every  y seconds)  and a retry count (for a Maximum of
z Retries).  The only rule of thumb is that  the retry counts should not be
excessive  (I like 3), and the "poll every x seconds" (the time-out value)
should be reasonable, (I like 2 sec  or more for remote nodes, reserving
sub-second values for nodes on the local subnet)  and the value of (
time-out   *  retries) is well-within the polling frequency, which is to
say that (y  * z) should be much less than x, so that netmon can give up on
a nonresponsive node and move on to another.  I hope you follow.

You say you are monitoring 900 devices all with SNMP enabled and your
ovobjprint shows about 24K objects, which is usually well within netmon's
capability.    So I'm at a loss to see where there might be a problem so
far.  yet you say that demandpoll is nonresponsive.   Are you using a seed
file?  Anything unusual about it?


James Shanks
Level 3 Support  for Tivoli NetView for UNIX and Windows
Tivoli Software / IBM Software Group


                                                                           
             "Skokan, Paul"                                                
             <Paul.Skokan@neta                                             
             pp.com>                                                    To 
             Sent by:                  <nv-l@lists.us.ibm.com>             
             owner-nv-l@lists.                                          cc 
             us.ibm.com                                                    
                                                                   Subject 
                                       RE: [nv-l] Polling Question         
             06/05/2006 03:38                                              
             PM                                                            
                                                                           
                                                                           
             Please respond to                                             
             nv-l@lists.us.ibm                                             
                   .com                                                    
                                                                           
                                                                           




Netview 7.1.4 FP4 on Windows 2003 platform.  Hardware is dual processor
XEON 3.6GHz with 4GB memory.  Polling a little under 900 network devices
(all snmp enabled).  I need to poll some devices more frequently than
every 20 minutes.

Here is ovsuf snip...
0:ovwdb:\usr\ov\bin\ovwdb.exe:OVs_YES_START::-O
-n40000:OVs_WELL_BEHAVED:120::

H:\>ovobjprint -S
Number of objects defined in the database: 24414

Total number of fields defined in the database is: 272.

Total number of field values in the database: 221169
         Number of Integer fields: 67968.
         Number of Boolean fields: 70829.
         Number of String fields: 67760.
         Number of Enum fields: 14612.

Thanks for you help.

Paul


-----Original Message-----
From: Evans, Bill [mailto:Bill.Evans@hq.doe.gov]
Sent: Monday, June 05, 2006 12:24 PM
To: nv-l@lists.us.ibm.com
Subject: RE: [nv-l] Polling Question


Hey, Paul; Welcome Newcomer. We need to know what you're running since
there are various flavors of NetView and they vary in performance
depending on the underlying hardware and operating system.

I've usually found the default is well configured if you have sized and
configured the system according to the suggestions in the NetView books.
For Windows it's a twenty minute polling cycle and for Unix/Linux it's a
five minute default polling cycle. Performance can be affected by a
number of other configuration variables as well as the total load in
terms of objects.

In particular the number of devices shown by the "ovobjprint -S" and the
number shown for the "-n" value on the ovwdb line in the
/usr/OV/conf/ovsuf line are critical (Unix/Linux file notation).  The -n
value MUST be a good percentage bigger.  If NETMON (the poller) is so
heavily loaded you can't sneak in a demand poll there's something out of
kilter in the configuration and the ovwdb -n value is the usual first
suspect.

Bill Evans

-----Original Message-----
From: owner-nv-l@lists.us.ibm.com [mailto:owner-nv-l@lists.us.ibm.com]
On Behalf Of Skokan, Paul
Sent: Monday, June 05, 2006 2:46 PM
To: nv-l@lists.us.ibm.com
Subject: [nv-l] Polling Question

I am fairly new to Netview and I need some advice on polling so that my
server can accomodate the # of devices I am monitoring...

I am primarly monitoring network devices (routers, switches, wireless,
etc).  The workload on the poller is heavily loaded at all times and it
is preventing me to do demand polls efficiently when troubleshooting
problems.  What is the best way to control the polling and SNMPcollect
scripts so that I can effectively monitor all devices at various levels?

Paul